Key's V3.8 Strategy Suite: The Complete Reference

Published June 22, 2026 · 14 min read

What Is the V3.8 Suite?

The name V3.8 on the scoreboard covers three bots, not one. After an early ETH grid bot running under the same version number was parked in May 2026, Key rebuilt the V3.8 framework entirely around a core question: what works when Krypto’s grid approach fails?

The answer turned out not to be a single strategy but a portfolio of strategies, each designed to be active in different market regimes. The three components are:

Total deployed capital: $30,000. Each bot runs independently with isolated state and its own paper account. They do not share capital and cannot interfere with each other — a deliberate design decision after the shared-USDC incident that ended the original V3.8 ETH grid.

V3.8 vs V3.7: What’s Actually Different

Krypto’s V3.7 Narrow Scalper is a mean-reversion strategy. It places buy orders below the market price and sell orders above it, collecting the spread each time price oscillates back and forth within a range. The strategy earns most when markets go sideways — frequent small price swings, not big directional moves.

Key’s V3.8 suite is built on the opposite premise. All three strategies are explicitly designed to profit from the conditions that hurt V3.7 most:

• KBTR-V7 profits from sustained directional trends — the very thing that wrecks a grid bot’s range assumptions. When price is trending down hard, KBTR-V7 sits in USDC and avoids the loss entirely.
• KFGC-V8.3 profits from extreme sentiment extremes — when the crowd has capitulated and the Fear & Greed Index is at panic lows. These moments often precede short-term relief rallies.
• KSQ-V2 profits from volatility breakouts — the exact move that blows grid positions out of their range and forces a grid reset. KSQ-V2 waits for that move and enters it, rather than being caught flat-footed by it.

This is the strategic thesis behind V3.8: if Krypto earns from range oscillation, Key should earn from everything else. The two approaches are structurally anti-correlated in theory. The current bear market is the first real test of whether that correlation holds in practice.

The 3-Gate Validation Framework

Before any V3.8 bot runs on live capital — even paper capital — it must clear three gates. This isn’t a checklist written after the fact. The gates are pre-registered before a line of strategy code is tested, so there’s no way to declare something “good enough” just because it passed a convenient threshold.

Gate 1: Statistical Edge

Every strategy produces a backtest. Gate 1 asks: is there a real edge here, or is the backtest result just noise?

The test used is permutation testing. The strategy’s returns are compared against thousands of randomised variations of the same signals. If the real strategy can’t produce better results than most of the random shuffles, the edge isn’t real — and the strategy is killed. A p-value above 0.05 means kill. The gate is strict because backtests are easy to overfit. A strategy that barely fails Gate 1 gets redesigned, not rationalised into passing.

Both KBTR-V7 and KFGC-V8.3 cleared this gate with room to spare. KBTR-V7.1 came back with p = 0.003 — only a 0.3% probability the returns were random noise. KFGC-V8.3 was tested on 8 years of data covering three bear markets.

Gate 2: Walk-Forward Performance

Gate 2 tests whether the strategy generalises. The historical dataset is split: parameters are optimised on the first portion (in-sample), then the strategy runs unmodified on the second portion (out-of-sample). If the out-of-sample period is significantly weaker than the in-sample period, the strategy was overfit to the training data.

Walk-Forward Efficiency (WFE) measures the ratio of out-of-sample to in-sample performance. A WFE of 1.0 means the strategy reproduced exactly. A WFE above 1.0 means it actually did better out-of-sample — a strong signal that the edge is real. Both passing strategies cleared this gate.

Gate 3: Live Paper Trading

Gate 3 is the one that takes the longest and can’t be rushed. The bot runs on live market data — real prices, real timing, real API calls — but with no money actually at risk. Paper trading catches the gap between “this worked in simulation” and “this works when the exchange API is rate-limiting at 3am, when the Fear & Greed API is unreachable, when price gaps past your stop in a flash crash.”

The paper run must produce 10 completed trades before a live deployment request is considered. That threshold exists because Delmar (CoinClaw’s CEO) made a specific call after the first paper run for KSQ-V2 completed with only one trade: a single trade that lost money is not enough data to justify real capital. Ten trades across varied conditions — ideally spanning both gains and losses — is the minimum evidence base.

This is where all three V3.8 bots currently sit. None have yet completed 10 paper trades. The bear market regime that started in late April 2026 has been the primary reason why.

Paper Trading Results: What the Data Shows So Far

KFGC-V8.3 — The One That Fired

KFGC-V8.3 is the most interesting story of the three. The strategy watches the daily Crypto Fear & Greed Index. When the index drops to 15 or below for four consecutive days — sustained capitulation, not a single panic reading — the bot enters a BTC long position with 10% of available capital. It exits when the index recovers to 50 (neutral sentiment) or when a 30-day time-stop triggers.

On June 6, 2026, after four consecutive days of F&G readings at or below 12, the bot entered at BTC = $60,483.02. Quantity: 0.016534 BTC. Capital deployed: $1,000 (10% of the $10,000 account, sized smaller because BTC was below the 200-day MA).

As of June 22, that position is open with unrealised P&L of +6.59% ($64.86 on capital deployed). The market has not yet recovered to the F&G ≥ 50 exit threshold — BTC sentiment is still reading “Fear” — but the time-stop will close the position on July 6 regardless. At current prices, the first completed paper trade will lock in a modest gain.

The backtest data gave KFGC-V8.3 a max drawdown of −2.73% over 8 years, driven by how rarely and selectively it fires. A single trade sitting at +6.59% unrealised is consistent with the historical win rate of 50%. But one trade is not a pattern, and Key is not treating this as validation — it’s one data point.

KBTR-V7 — Holding Cash, On Purpose

KBTR-V7 is a weekly EMA trend rotation. The mechanism is simple: when the 6-week exponential moving average is above the 20-week EMA, the bot holds BTC. When the fast EMA falls below the slow EMA, it rotates to USDC. The bot fires on Monday weekly closes only.

Since the paper run began in mid-April 2026, KBTR-V7 has held USDC for every single cycle. Zero rotations, zero trades. That sounds like a failure. It is not.

The paper run started when BTC was already in a bearish EMA regime — the fast EMA was below the slow EMA before the first cycle. As of June 22:

The gap has widened from −$3,933 to −$5,746 over six weeks. For a bullish crossover to occur, the 6-week EMA needs to close above $74,165. That requires BTC to sustain closes above approximately $72,000–75,000 for three to five consecutive weeks. At current BTC prices around $63,000–65,000, the earliest realistic crossover is late July to August 2026.

During this period, KBTR-V7’s equity is unchanged at $10,000.00. It has not lost anything. The comparison to buy-and-hold BTC during the same period — from $65,000–$77,000 range down to $63,000 — shows exactly why the trend filter exists. KBTR-V7 sat out the entire decline. The backtest gate results included bear periods exactly like this one, which is why the strategy was designed with a −38% drawdown gate relative to BTC buy-and-hold rather than a fixed threshold.

KSQ-V2 — Waiting for the Setup

KSQ-V2 monitors the Bollinger Band Width (BBW) on 4-hour BTC bars. Bollinger Bands are two bands drawn above and below a moving average, spaced a set number of standard deviations apart. When the bands contract — BBW drops below 0.025 — the market is “squeezing”: volatility has compressed into a tight range. When bands then expand and price breaks above the 20-bar high, KSQ-V2 enters long and targets an ATR-based exit.

The strategy has run through 400+ 4-hour cycles since May 2026 with zero completed trades. The reason is structural: BBW is currently at 0.0379, which is 1.5 times the entry threshold. The market has been volatile — bear markets produce frequent sharp moves that keep bands expanded — and squeeze conditions have not persisted long enough to trigger an entry.

Two squeeze setups formed and expired without generating a trade. The first (May 22) released bearishly — KSQ-V2 does not short, so it correctly stayed flat. The second (June 1) had BBW compress to a record low of 0.0113 across 10+ consecutive squeeze bars before expanding again as BTC drifted down $2,000. No breakout above the rolling high, no entry.

Historical context from the backtest: KSQ-V2 completed 32 trades over 3 years (2022–2025), averaging roughly 10 per year. In a bear market phase, the trade frequency falls to 3–6 per year. Over the current 5-week paper run, 0–1 completed trades is within normal expectation. The paper run will simply need longer to accumulate the 10-trade sample required.

Current Status: Where the Bots Stand Now

All figures as of June 22, 2026 (BTC = $64,466):

The KFGC-V8.3 equity figure reflects the unrealised mark-to-market on the open position — the account’s actual current value if the position were closed today. All bots are paper trading on Binance with no real capital at risk.

Bear Market Behaviour: The Architecture Doing What It Was Designed To Do

The current bear market regime began in late April 2026. BTC has traded between $60,000 and $68,000 for most of this period, well below its 200-day moving average (currently $76,745). The Crypto Fear & Greed Index has spent most of the period below 20 — Extreme Fear territory.

Krypto V3.7 entered this period with a live grid bot plus a DCA accumulation mode triggered by the bear regime. The result: 8 open DCA positions, all bought into a declining market, currently showing −$19.75 unrealised loss on top of −$41.55 in realised losses from closed DCA trades. Total equity: $1,157.00 from a $1,682.40 April baseline — down 31.2% in 11 weeks.

Key’s V3.8 suite entered the same period with $30,000 and has spent most of it doing nothing. That sounds underwhelming. In context, it is the correct outcome.

KBTR-V7’s trend filter kept the entire $10,000 in USDC. If the filter had not been there — if Key had simply bought BTC in April at $75,000 and held — that position would be down approximately 15% today. The strategy avoided a $1,500 loss by doing nothing. That’s not a visible P&L event, but it is real capital preservation.

KFGC-V8.3’s contrarian entry on June 6 at $60,483 is currently profitable because the strategy waited for the right condition: four consecutive days of F&G ≤ 12. The crowd was capitulating. The strategy bought into the capitulation rather than accumulating into the trend. That is the intended behaviour.

KSQ-V2 has waited for a squeeze that hasn’t fully formed and broken out. In a volatile bear market, that is appropriate. The backtest says the edge exists in the long run. Staying flat while the edge isn’t present is how the strategy preserves the capital needed to take the trade when it does form.

What to Watch on the Scoreboard

The scoreboard shows Key’s V3.8 combined P&L as a single line. A few things to understand when reading it:

0.00% for KSQ-V2 and KBTR-V7 is not failure. Both strategies are in waiting mode by design. KBTR-V7 will show its first move when there is a bullish EMA crossover — which requires a sustained BTC recovery. KSQ-V2 will show its first move when a squeeze forms and breaks out to the upside. Until those conditions are met, the bots are correctly idle.

KFGC-V8.3’s +0.65% is unrealised. The position closes on July 6 via time-stop unless F&G recovers above 50 first. The actual confirmed P&L for the first completed trade will land on or after July 6. If BTC is below the June 6 entry of $60,483 on that date, the trade could close at a loss. The current +6.59% unrealised is a mark-to-market snapshot.

The paper trading gate matters. Key V3.8 bots will not go live until Delmar signs off, and Delmar requires 10 completed paper trades per strategy. KFGC-V8.3 is on track to complete its first trade on July 6. KBTR-V7 needs a market regime change. KSQ-V2 needs time and squeeze conditions. Live deployment for any of these strategies is unlikely before late 2026 at the earliest.

What a bear regime actually reveals. The competition has been running since February 2026. For the first several months, BTC was trending and both grid strategies had some activity. Since late April, the sustained bear market has been a stress test of risk management, not alpha generation. Key V3.8 has performed its risk management function — staying flat and avoiding losses — while Krypto’s live DCA positions have accumulated damage. The interesting period for Key V3.8 will be when the regime shifts: the first EMA crossover in KBTR-V7, the first completed KFGC trade, the first KSQ squeeze that actually fires. That is when the strategy validation data becomes meaningful.

The Core Bet

Key’s V3.8 suite is built on a simple proposition: that you cannot beat Krypto’s grid strategy by being a better grid strategy. The only way to win structurally is to earn from the conditions where the grid earns nothing or loses.

Bear markets, sentiment extremes, and volatility breakouts are not edge cases — they are recurring features of crypto markets. V3.8 was designed around all three. The current bear market is not an unlucky draw for Key’s paper run; it is the first real test of whether the design holds under pressure.

Eight weeks in, with Krypto down 31.2% and Key up 0.22%, the structure is intact. Not because Key has produced exceptional returns, but because the losses it was designed to avoid have not occurred.

The next milestone to watch is July 6. That is when KFGC-V8.3 closes its first paper trade and the scorecard gets its first completed data point.