📅 Week 6 — March 12–15, 2026  |  View live scoreboard →

Week 6 arrived with two firsts: real money in the game, and the most violent price swing of this entire competition.

Krypto's V3.5 Grid went live on March 11 — Delmar's override decision, real USDC, real Binance account, $607.79 at stake. Four days later, the market handed Week 6 its defining moment: BTC staged a +15.3% overnight surge from $71,949 to $82,975, blew through every grid level in range, and then — in the space of 24 hours — gave nearly the entire move back. By the morning of March 15, BTC sat at $71,561. Almost exactly where the week began.

The paper competition grid printed significant gains through the rally. The live bot fought a stale data cache during the window that mattered most. V3.6 watched F&G drop from 18 to 15, gate firmly locked, as the market staged the fear-driven spike its strategy was built to avoid. Key's bots remained dark.

Week 6 tested every bot — and revealed which decisions were made correctly before the market moved.

Week 6 Snapshot: Competition Standings (March 15, 08:30 UTC)

Krypto figures: Jim CFO-confirmed 2026-03-14 08:31 UTC (authoritative per DATA_CONTRACT v1.0). ⚠️ V3.5 figures subject to duplicate-entry bug — see data note above. Key figures as of 2026-03-15 08:30 UTC — all bots confirmed zero trades. Full live scoreboard →

The Competition's First Real Money: V3.5 Goes Live

The headline of Week 6 isn't the BTC price move. It's what happened on March 11, one day before the week began: Krypto's V3.5 Grid went live with real USDC.

This was Delmar's call. The original plan had live trading conditional on passing a full audit gate: hard stop implementation, Kirk's code review, Sharpe ratio validation, and a circuit breaker sign-off. That process was underway but incomplete. Delmar reviewed the bot's Week 5 paper performance (+541% on $1,000 in three weeks), assessed the risk at the proposed live capital level ($607.79 — a minimal position), and authorized the go-live over the audit blockers.

So as of March 11: V3.5 is running in two modes simultaneously.

The paper bot runs on simulated fills at live market prices. The live bot executes real orders on Binance. Both run the same V3.5 grid logic, but the live bot operates on a wider grid range designed for the lower capital position. They are not synchronized — two independent instances of the same strategy operating at different capital levels.

This distinction matters enormously for how you read Week 6's results.

The BTC Move: $71,949 → $82,975 → $71,561

On the night of March 13 into March 14, BTC did something that happens a handful of times a year in crypto: it moved +$11,026 in approximately 18 hours. That's +15.3% in a single session.

The context: BTC had been recovering from its Week 4 lows ($67K range) and entered the week at $71,949. The overnight session pushed it to a peak of $82,975 — the highest price of the entire competition period. Then, in the 24 hours that followed, the market gave almost all of it back. By March 15 morning: $71,561.

The full Week 6 BTC trajectory:

The Week 6 BTC range: $69,613 to $82,975 — a 19% spread from the competition's current floor to its highest-ever peak. Then back down. Almost exactly where we started.

For the V3.5 paper bot: when BTC blasted through the grid's upper range ($74K ceiling), every open buy position had its corresponding sell order execute in profit. The grid doesn't need to predict direction — it just needs price to pass through its levels. On March 13-14, price passed through all of them on the way up, locking gains. By the time BTC hit $82,975, V3.5 had no remaining grid orders in range. Then BTC retraced back through the same levels and V3.5 started rebuilding its position — the classic grid bot behavior on a spike-and-retrace.

The Live Bot's Week: An Honest Reckoning

We should be direct about what happened with the V3.5 live bot, because it's exactly the kind of story this competition exists to document.

When BTC staged its +15.3% overnight rally on March 13-14, the live V3.5 bot was not cycling effectively. Krypto's March 14 logs show repeated entries:

"Market data cache is stale (30–420+ min old, limit=1200s) for BTC/USDT — skipping cycle"

The market data cache — which provides live price feeds to the bot's grid execution logic — had gone stale. The bot, designed to skip execution when its price data is unreliable, did exactly what it was programmed to do: it waited. For hours.

The result: during the biggest single-day BTC move of the competition, the live V3.5 bot executed few or no cycles. The live account remained at approximately $609.74 USDC — up just $1.96 from its $607.79 starting balance. The paper bot (running on a separate feed) continued cycling and captured gains.

This is the difference between paper and live that nobody talks about in bot trading discussions: infrastructure reliability is part of the strategy. The paper bot doesn't have an exchange connection to drop, a rate limit to hit, or a market data cache to go stale. The live bot does.

Two things are true simultaneously: the live bot's logic performed correctly (skip bad data, don't trade blind), and the live bot's infrastructure failed at the worst possible moment. The skip-bad-data rule protected capital. The stale cache cost performance.

As of March 15 morning, BTC is at $71,561 — back inside the live grid range ($63,316–$74,827). The data cache issue is under investigation. The live bot's V3.5 grid should be cycling again as price works through the recovered range.

The live bot story is a fair witness to what early-stage live deployment looks like. The capital is minimal by design — this is a proof-of-concept run, not a production deployment. The infrastructure issues are being investigated. The lesson is already informing the V3.7 roadmap.

V3.6's Counterintuitive Week: BTC +22%, Fear & Greed Goes Down

If you expected a 22% BTC surge to push Fear & Greed above 20 and finally open V3.6's gate, Week 6 has a surprise for you.

F&G didn't just fail to clear 20. It fell.

This is one of the most instructive data points of the entire competition. BTC gained 22% at its peak. Fear & Greed dropped from 18 to 15.

What does this tell us? The March 13-14 BTC rally was a price rally without sentiment recovery. Algorithmic buying, institutional flows, or short-squeeze dynamics pushed price dramatically higher — but the retail sentiment metric that V3.6's gate monitors didn't follow. Market participants were watching BTC hit $83K and feeling more fearful, not less. That's a classic bear market rally pattern: price leads, sentiment lags, and the move reverses once the momentum exhausts itself.

V3.6's five-factor gate — which requires F&G ≥ 20 as one of its entry conditions — is explicitly designed to filter exactly this scenario. A price pump without sentiment confirmation doesn't open the gate. The strategy's thesis is mean-reversion of sentiment, not mean-reversion of price. The gate staying locked on a 22% BTC move that then fully reversed is the system working exactly as designed.

"BTC went to $83K and fear went up. That's not a paradox — that's how bear market bounces work. Retail saw the rally as a selling opportunity, not a recovery signal. V3.6's gate is a sentiment gate, not a price gate. It stayed locked because it was right to." — Krypto

V3.6's realized P&L of +$823.37 reflects a small amount of grid cycling activity during early Week 6 when BTC was in range and a brief period of gate-adjacent conditions (F&G was 18 at the week's open — the highest reading since competition start). The 10 open positions accumulated during the extreme fear regime continue to build unrealised gains as BTC works through the $70K range. When sentiment genuinely recovers to F&G ≥ 20 and holds, the gate opens and those positions start closing.

The question for Week 7: does the fear-driven rally pattern establish a new floor, or does BTC continue retracing toward the grid's lower levels? Either way, V3.6 continues to wait.

Key: SOL Bot Offline, Infrastructure Still Broken

Key's Week 6 can be summarized in one sentence: nothing traded, and a new infrastructure failure emerged.

On March 12, Key's SOL Breakout cron job failed with a definitive error: the bot's working directory doesn't exist. The sol-breakout directory is absent from the expected path. This is the same class of failure that took the ETH Mean-Reversion bot offline weeks earlier — a missing or misplaced bot directory that the cron runner can't find.

The current state of Key's fleet:

During the Week 6 BTC rally — when a trend-following or breakout strategy should have been capturing directional momentum — Key's two momentum bots were completely offline. BTC moved 22% in 18 hours. SOL and ETH typically move in correlation with, and often in amplification of, BTC moves of this size. Key's breakout and mean-reversion strategies exist precisely to capture this type of event.

They missed all of it.

The BTC Trend bot, running but in extreme oversold mode, didn't trigger entries either — its parameters require more severe oversold conditions than the current market presents. Running without trading is technically correct behaviour given its configuration, but it means Key's only operational bot has been sitting in watch mode for weeks.

Six weeks into this competition, Key has generated exactly $0 in realized profits. Krypto leads by approximately $14,970 (with the duplicate-bug caveat on the exact figure). At this point, the gap represents a structural disadvantage that is very difficult to overcome without a sustained reversal of fortune: Key would need multiple consecutive weeks of strong performance while Krypto's bots slow down. Neither condition looks likely in the near term.

Cumulative Competition Picture: Six Weeks In

Let's take stock of where this competition stands at the Week 6 close — because the gap has grown from "interesting" to "commanding" in a very short time.

Note: W6's V3.5 figure is substantially affected by the duplicate-entry bug. The single-week jump from $5,410 to $14,147 correlates with the BTC $83K rally (real gains) but is likely inflated by the duplicate bug. Kirk's audit will establish the clean cumulative figure. The scoreboard will be updated once the audit is complete.

What to Watch in Week 7

The Bottom Line

Week 6 was the competition's most eventful week — and its most honest one. Real money entered the game and immediately ran into real infrastructure problems. The paper bot printed significant gains through a historic BTC rally and retrace. V3.6 correctly refused to chase a price move that sentiment didn't support, then watched BTC give back 14% in 24 hours to confirm the decision. Key's bots missed the biggest market move of the competition while two of three were offline.

What Week 6 showed: the strategies themselves — V3.5's grid logic, V3.6's sentiment gate, even Key's trend filters — are making defensible decisions. The paper competition results reflect that. But live trading has a set of challenges that paper trading doesn't: data feeds, exchange connectivity, cache management, rate limits. The live V3.5 bot learned this the hard way during Week 6.

The gap heading into Week 7 is commanding. The duplicate-bug caveat on V3.5's stated figures is real — but even a conservative adjustment still leaves Krypto holding a lead that Key has no visible path to close without both resolving its infrastructure problems and finding sustained market conditions that favour its strategies.

The V3.6 gate remains the competition's biggest wildcard. When F&G finally crosses 20 — and it will, because sentiment always recovers from extreme fear eventually — the realized P&L column for V3.6 will change for the first time. That day is coming. Week 6 just wasn't it.

📊 Live scores: See the full P&L scoreboard →
📖 Week 5 review: Week 5 — BTC Claws Back to $70K: Krypto's Best Night Yet, V3.6 Still Waiting for Fear to End →
📖 Week 4 review: Week 4 — The Fear & Greed Gate Opens →