Week 17-18: V37 goes live into a rising market

On April 24, Krypto deployed $1,682.40 into the V3.7 Narrow Scalper - the first real-money live bot in the competition. BTC was trading around $77,000 at the time. The scalper works by holding BTC and selling when price spikes above a tight range, then waiting to buy back cheaper. In a stable or slowly rising market, it grinds out small wins. In a declining market, it wins the trades but loses ground to the underlying.

The first two weeks went well. Five trades in week 17 (April 25-26), one in week 18. All six closed green. The largest single win was $0.34 on a sell at $77,609. None of these numbers look impressive in isolation - the bot targets $0.10-$0.35 per trade - but the key metric at this stage was consistency. Every trade had a positive profit. The win rate was holding at 100%.

Key, meanwhile, had no live bots active. The paper trend bot from weeks 11-16 was still accumulating paper results, but nothing was running with real capital.

Weeks 19-20: the busiest stretch

The period from May 4 to May 15 was V37’s most active. Ten trades in week 20 alone. BTC was still in the $74K-$77K range and vol was high enough for the scalper to find entries. Total realized P&L for the month: around +$1.60.

Week 20 (May 11-15) had the densest trade activity: trades on May 11, 12, 13, 14, and 15 - thirteen entries and exits in five days. The strategy was working as designed. Each trade captured a small slice of volatility. None lost money.

Then BTC started falling. Not a crash, just a grinding slide. From $77K in late April, it drifted toward $70K, then $68K, then kept going. The scalper’s DCA mechanism started buying the dips - the idea being that if BTC pulls back temporarily, you average down and sell the recovery. By early June, the bot had deployed $396.98 across eight DCA positions averaging $67,173 per BTC (buys ranged from $66,288 to $68,541 across June 2–3). At the time, that looked like a reasonable average-down price.

Weeks 21-24: five weeks of silence

From around May 19 to June 15, V37 made zero trades. Not one. The regime filter was doing its job - the bot detected a sustained bear trend and blocked all new entries. This is by design: the narrow scalper isn’t supposed to keep firing in a trending bear market, because each sell-and-rebuy would just lock in losses at successively lower prices.

Five weeks is a long time to sit still. BTC fell from roughly $70K to $63K during this period. The DCA positions bought at an average of $67,173 were now underwater. The USDC portion of the portfolio was fine - no active positions meant no active losses. But the BTC held had declined in value as the price moved against it.

From the outside, it looked like the bot had stopped working. It hadn’t. Staying flat in a trending bear is the correct move for a narrow scalper. The problem is that “correct move” and “portfolio value increasing” can be different things.

Weeks 25-26: the bot comes back

The bot resumed trading on June 16. BTC had stabilized around $62,000-$64,000 and the regime filter cleared. Eleven trades between June 16 and 21 - the second-busiest stretch behind week 20. All profitable. The current BTC price of $62,548 is below the average DCA buy price of $67,173, meaning those positions are still underwater. But the bot itself hasn’t lost a single closed trade.

As of June 23: 35 trades closed, 35 wins. Realized P&L from trading: +$2.65. Portfolio value: ~$938 against a starting balance of $1,682.40.

That gap — $2.65 realized profit against a ~$744 portfolio decline — tells the whole story of the past ten weeks. The scalper won every trade. The market moved against the BTC exposure faster than small scalp wins could offset it.

V3.8 ETH grid: 15 trades, then parked

Krypto also ran a second live bot during this period: V3.8 ETH Grid, deployed May 4 with $1,964.35. The full first-week story is in V3.8 ETH grid: the first week. The condensed version: 15 trades between May 7 and May 14, $13.68 total profit, 100% win rate.

On May 12, Delmar made the call to park V3.8. Buy orders were cancelled, ETH positions were drained via sell orders, and the freed USDC was redirected to V3.7 for its Option D grid recentre. The bot still holds 0.211 ETH (~$351 at current prices). It has not traded since May 14.

That parking decision is part of what the combined P&L number captures. V38 made $13.68 in 8 days of trading, then sat still. V37 ground out $2.65 across 35 trades over two months. Together: $16.33 realized on $3,646.75 starting capital.

Key’s situation

Key has had no live capital at risk during the entire ten-week period. No live trades. No losses. No gains. The paper bot work continued in the background - research into RPA V1 (a risk-parity allocation strategy), FGC V8.3, Trend V7, Squeeze V2 - but none of them graduated to live capital during April, May, or most of June.

On June 22, the RPA V1 paper bot launched. The strategy allocates capital across tranches, with the first two tranches at 0.1 BTC each (total $10,000 paper capital). Zero trades so far. The strategy is in its first hours of running.

Krypto has had two live bots running for ten weeks. Key has been paper-trading the entire time. That was Krypto’s advantage. Whether it turns into a lead depends on where BTC goes. If price recovers above $67K, Krypto’s DCA positions recover and V37’s 100% win rate looks like the right strategy all along. If BTC stays flat or falls further, Key’s patience starts looking like discipline rather than delay.

What to watch next

Three things worth tracking:

$67,173 is Krypto’s average DCA entry. If BTC recovers above that, the portfolio loss narrows fast. That’s the line between V37 looking unlucky and V37 looking like it worked.

RPA V1 paper results will be the most interesting thing to watch on Key’s side. If the paper run is strong over the next 30-60 days, Key will almost certainly push to go live. That’s the first point at which the competition becomes properly two-sided again.

The competition hasn’t swung decisively either way. Krypto is in a hole on paper but has the experience of running live bots for ten weeks. Key has no losses, no gains, and the data from all those paper runs. Both of those things matter.

Next update when there’s a material change — a V37 losing trade (which would break the 35-trade streak), V3.8 coming back off the bench, or Key going live.