What Actually Worked in 2026?

In H1 2026, win rate climbed in a straight line from GRID (35.14%) to spot DCA (40.95%) to DCA Futures (50.92%) to COMBO (57.85%) — a 23-point gap between the least and most adaptive bot type on the platform, in a market that spent more of the year trending than ranging. COMBO and DCA Futures also closed positions roughly 3.7–5x faster than GRID and beat it on every return measure. The one number worth reading carefully rather than repeating is average return: COMBO's 15.77% average sits nearly 9x above its own 1.76% median, meaning a handful of large wins pulled the average up far past what a typical closed bot actually earned.

TL;DR

• Win rate rose with adaptiveness: GRID 35.14% → DCA 40.95% → DCA Futures 50.92% → COMBO 57.85%
• COMBO and DCA Futures beat GRID and spot DCA on every metric measured: win rate, average return, median return, and time to close
• Spot DCA's average return was slightly negative (-0.1%) despite a barely-positive median (0.01%) — a small number of larger losers dragged the average below what most closed bots actually did
• Average return figures are heavily skewed by outliers across all four bot types (4.6x–14.2x mean-to-median gap) — median is the more representative number for "what a typical bot did"
• COMBO and DCA Futures closed positions in roughly 2–3 days on average, versus 8.8–10.2 days for spot DCA and GRID
• Higher win rate and return on the leveraged bot types come with higher liquidation risk — this report covers performance, not risk-adjusted return, see the risk note below

H1 2026 at a Glance

Here's the full H1 2026 performance snapshot for each crypto trading bot type on Bitsgap, side by side, before the walkthrough of what's behind each number:

H1 2026, Bitsgap internal platform data. Win rate = % of closed bots that closed in profit, regardless of position size.

A Note on How to Read This Data

This report uses two different return numbers side by side on purpose: average return and median return. They tell different stories, and collapsing them into one headline number would misrepresent what most traders actually experienced.

Average return is pulled upward by a small number of large winners — a bot that returns 200% on one favorable trend has an outsized effect on the average across a few hundred bots. Median return describes the bot in the exact middle of the distribution: half of closed bots did better, half did worse. When the two numbers are close, the average is a fair summary. When they're far apart, as they are here, the average describes what the best outcomes achieved, not what a typical trader should expect.

This report treats median as the more honest answer to "what actually worked," and shows average alongside it so the gap itself — and what it implies about outlier-driven, leverage-amplified outcomes — is visible rather than hidden.

Win rate carries a similar limitation, in a different direction: it counts any closed bot that ended above zero as a "win," whether that bot closed at +0.01% or +40%. A high win rate says a strategy closed in profit often — it says nothing about how large those profits typically were. That's why this report always reads win rate next to median return rather than on its own; a bot type can have a strong win rate and a modest typical return at the same time, and both numbers are needed to see that.

One more gap between theory and this table: none of these figures isolate execution cost. The price a bot's order targets and the price it actually fills at aren't always the same number, and that gap eats into realized return on every single trade, not just the ones that lose.

The Market GRID Was Built For vs. The Market It Got

A GRID bot profits from price oscillating inside a range — it buys low, sells high, and repeats, without caring which direction the range eventually breaks. That mechanic needs one condition to work: the market has to actually range. When a market trends persistently in one direction instead, a grid keeps buying into a move that never reverses back through its range, and the strategy that thrives on chop starts absorbing a one-way loss instead.

H1 2026 leaned toward the second scenario for a meaningful stretch of the year, and the performance data lines up with that directly: GRID posted the lowest win rate of the four bot types (35.14%) and the slowest average time to close (10.23 days) — consistent with positions that sat open waiting for a range that didn't fully materialize, rather than cycling through quick, decisive closes.

Where COMBO and DCA Futures Actually Pulled Ahead

The launch data (see Sources) already showed COMBO and DCA Futures picking up share as the year went on. The performance data — in the table above — confirms that shift was justified, not just sentiment-driven.

COMBO's win rate is 22.7 points higher than GRID's, meaning it closed in profit roughly 1.65x as often over the same stretch. Speed tells a similar story: DCA Futures and COMBO closed in 2–3 days on average, against 8.8–10.2 days for spot DCA and GRID, so the leveraged strategies weren't just winning more, they resolved faster too, which matters for capital efficiency on its own. Spot DCA is the outlier here. Its average return went slightly negative (-0.1%) even though the median stayed barely positive (0.01%); most spot DCA bots roughly broke even, and a handful of larger losses pulled the average below zero.

COMBO's structural advantage in this window comes from layering GRID's range logic and DCA's averaging inside one leveraged futures position — it isn't fully dependent on the market ranging the way a standalone GRID bot is. In a year that trended more than it ranged, that flexibility shows up directly in the numbers. One trader's account shows this at the individual level: three leveraged COMBO Shorts alongside one unleveraged spot DCA bot, run through a falling market in July 2026.

The risk this data doesn't show. COMBO and DCA Futures win on every performance metric in this table, and leverage is a direct driver of that. It's also why the mean-to-median gap is so wide, and why an adverse move can end in liquidation rather than an open position waiting to recover. Spot GRID and spot DCA bots can sit at an unrealized loss without a forced close. A leveraged bot at the same paper loss can be liquidated outright, ending the trade regardless of what price does next. This measures realized performance on closed bots, not liquidation risk — higher win rate or return numbers on a leveraged bot type reflect a different risk taken on, not a safer one. A capital allocation framework for running several bots without one leveraged position quietly becoming the whole account covers how to size around exactly this.

What "Actually Worked" Means Once You Include the Skew

COMBO didn't return 15.77% for a typical trader — that number describes the best-performing bots, not the median one. What actually held up: COMBO and DCA Futures beat GRID and spot DCA on every metric measured, by a wide and consistent margin, closing positions several times faster in the process. That holds whether you check win rate, median return, or average return — it doesn't depend on picking the most flattering number.

The skew itself is also information. A 9–14x gap between average and median on the leveraged bot types is a direct signature of leverage: it amplifies the good outcomes far more than the typical ones, which is exactly why median, not average, is the number to anchor expectations on.

This is also a snapshot of one specific market regime, not a permanent ranking of crypto trading bot types. GRID's numbers reflect a stretch of H1 2026 that trended more than it ranged, and a trending market is precisely what GRID's mechanic struggles with. A sustained ranging market would be expected to close that gap, and could plausibly reverse which bot type leads on win rate — conditions decide this ranking, not a fixed hierarchy.

FAQ

Which bot type had the highest win rate in H1 2026? COMBO had the highest win rate at 57.85%, followed by DCA Futures at 50.92%, spot DCA at 40.95%, and GRID at 35.14% — a 22.7-point gap between the highest and lowest.

Does a 15.77% average return mean COMBO bots typically returned that much? No. COMBO's median return was 1.76%, meaning half of closed COMBO bots returned less than that. The 15.77% average was pulled up by a smaller number of large-winning bots, which is common in leveraged strategies.

Did any bot type lose money on average in H1 2026? Spot DCA had a slightly negative average return (-0.1%) despite a near-zero positive median (0.01%), meaning the typical closed bot roughly broke even while a subset of larger losses dragged the average below zero.

Why did COMBO and DCA Futures close positions faster than GRID? COMBO and DCA Futures averaged 2.05–2.79 days to close versus 8.79–10.23 days for spot DCA and GRID. Both faster-closing types use leveraged futures positions, which tend to resolve trades more decisively than spot strategies waiting on a price range.

Why did GRID perform worst on win rate in H1 2026? GRID bots need a ranging market to work as designed. H1 2026 trended for a meaningful stretch of the year, which works against GRID's core mechanic and lines up with its lower win rate and longer average time to close compared to the other three bot types.

Does COMBO's higher win rate mean it's a lower-risk bot type? No. COMBO and DCA Futures use leverage, which is a direct driver of their stronger H1 2026 numbers but also means an adverse price move can trigger liquidation rather than leaving an open position to recover, unlike spot GRID or spot DCA. Higher performance in this data reflects different risk, not lower risk.

Will GRID and DCA always underperform COMBO and DCA Futures? Not necessarily. This report reflects H1 2026, a period that trended more than it ranged — conditions that specifically disadvantage GRID's range-based mechanic. A sustained ranging market would be expected to narrow or reverse this gap.