Bitcoin 
Ethereum 
BNB 
XRP 
Solana 
TRON 
Dogecoin 
Cardano 
Hyperliquid 
Zcash 
Chainlink 
Monero 
Stellar 
Sui 
Hedera 
Shiba Inu 
Bittensor 
Uniswap 
Polkadot 
World Liberty Financial 
Aster 
Ripple USD 
Render 
Artificial Superintelligence Alliance 
ETH Started Running. BTC Didn't Get the Memo.
BTC entered the July 18 backtest window under pressure — already tracking lower while institutional capital was quietly rotating into ETH staking narratives.
Then ETH broke out. Hard.
By mid-August, ETH had surged +24% while BTC simultaneously fell -7% — a combined 31-percentage-point spread between two assets that usually move together. For a spot holder with a 50/50 BTC/ETH portfolio, the blended return on passive holding was approximately +8.5%. Not bad for a turbulent month.
The question we wanted to answer: What does a rebalance bot do when its two assets stop correlating and start diverging violently?
We ran three parameter variants — different ratio thresholds and time triggers — across the full divergence window using real Binance 1-minute OHLCV data to find out.
Strategy Parameters
How Each Setting Impacted Performance?
Every parameter had a specific job in this setup. August’s violent divergence exposed what happens when the market refuses to cooperate with that job.
Parameter Impact Summary
| Parameter | Impact | The Logic (Why) |
|---|---|---|
| 50/50 Allocation | 📉 Drove portfolio loss | ETH crash outweighed BTC stability |
| 2% Ratio Threshold | 🔄 Triggered 3 sells of ETH | Sold winner early, missed full ETH run |
| By Coin Ratio Logic | ⚖️ Forced relative rebalancing | Systematic buying of falling assets |
| No Time Rebalance | 🛡️ Minimized fee drag | Reduced unnecessary BTC-to-ETH conversions |
| 0.1% Fee / Conversion | 🏁 Crystallized total loss | Fees negligible; end-date timing critical |
1 trade. $6.32 profit. Or 12 trades. $102.72 loss. Pick your threshold carefully.
📝 The math that matters
💰 The Bottom Line on Variant B: The only variant that avoided a loss returned just $6.32 on a $5,000 deployment — an effective monthly yield of 0.13%. Its single trade happened to hit a moment when the ratio briefly swung back toward parity. Everything after that was simply staying out of the way.
Annualizing 0.13% gives you 1.56% projected annual return, which a savings account would beat. The significance here isn’t the profit; it’s the survival. Variants A and C didn’t survive.
⚡ The Over-Trading Penalty: Variant C executed 12 swaps and lost $102.72. Variant A executed 3 swaps and lost $104.80. The math is brutal: more trades in a diverging market = more sells of the winning asset (ETH) and more buys of the losing asset (BTC). Variant C’s 12 trades systematically transferred capital from ETH’s 24% rally into BTC’s -7% decline — 12 separate times. The bot didn’t make bad decisions. It made the same rational but wrong decision 12 times, perfectly.
🛡️ The Passive HODL Advantage: A $5,000 investment split 50/50 in BTC and ETH, touched zero times across the backtest window, would have returned approximately +$425 — an 8.5% gain. Every rebalancing action moved capital away from that outcome. The divergence didn’t punish the strategy a little. It exposed rebalancing’s core mechanical weakness completely: in a trending divergence, the bot is programmed to fight the trend.
Here comes our A/B/C strategies quick comparison:
| Variant | Threshold | Trades | ROI % | P&L (USDT) |
|---|---|---|---|---|
| A | 2% | 3 | −2.10% | −$104.80 |
| BThis Playbook | 5% | 1 | +0.13% | +$6.32 |
| C | 1% | 12 | −2.05% | −$102.72 |
The pattern here is unambiguous: the wider the threshold and the less the bot traded, the better it performed. Variant B’s single trade was the closest the strategy came to neutral. Variant C’s 12 trades — driven by the tightest ratio trigger combined with a 30-minute time clock — converted a potential +8.5% HODL month into a -2.05% loss through sheer mechanical over-activity.
This is not a parameter optimization problem. It’s a market condition problem. No rebalance threshold would have beaten simple HODL in a sustained, directional ETH divergence. The lesson from Variant B isn’t “use 5% threshold always.” It’s “in trending divergence, the optimal rebalance frequency approaches zero.”
What the results are really telling you.
✅ what worked
Variant B’s 5% ratio threshold with no time trigger was the only configuration that stayed out of its own way. By requiring a 5-point allocation drift before acting — and removing the time-based override entirely — it fired exactly once and stopped.
That single swap likely caught a brief moment of ETH/BTC ratio compression during the divergence window, locking in a small positive spread. Fee management was clean: one swap on $5,000 notional at 0.1% costs approximately $5, leaving the $6.32 gross gain with minimal drag. The lesson is structural: passivity was the strategy’s only real edge here.
⚠️What didn't work
ETH’s +24% move never reversed back toward BTC during the test window. Every ratio-based rebalance trigger interpreted ETH’s outperformance as “drift” requiring correction — meaning the bot sold ETH (winner) and bought BTC (loser) each time the threshold was crossed.
Variant C did this 12 times. Each swap locked in ETH profits prematurely and added to a BTC position that kept declining. There was no mean reversion to harvest. The core mechanical assumption of ratio rebalancing — that divergent assets will converge — was simply wrong for this specific market event. The August 2025 ETH institutional inflow wasn’t noise. It was a structural move, and the bot couldn’t tell the difference.
💡 The key insight
Rebalancing is a mean-reversion bet. Every trigger assumes the ratio will converge again.
When ETH surged on genuine fundamental catalysts — $4B in institutional ETF inflows, staking narrative, real demand — the ratio didn’t converge. It kept diverging. The rebalance bot responded to each new divergence signal by selling more ETH and buying more BTC, each time with perfect logic and catastrophically wrong timing.
The real enemy isn’t volatility. It’s directional conviction in one asset.
Rebalancing harvests spread when assets oscillate around each other. When one asset breaks away from the other on fundamental news, the bot doesn’t adapt — it doubles down on the correlation assumption that just stopped being true. Before deploying a rebalance bot on any pair, the most important question isn’t “what threshold?” It’s: “Is there a fundamental reason one of these assets might decouple from the other this month?”
🚩 Watch out for - a potential red flag
The time-based trigger is the silent killer in this backtest.
Variant A’s combination of 2% ratio threshold AND a 1-minute time trigger created a double-fire condition: the bot could rebalance either when the ratio drifted 2% OR every minute — whichever came first. In a fast-moving divergence like August 2025’s ETH surge, the 1-minute clock essentially made it a near-continuous rebalancer. Three trades sounds low. But each one was large enough on a $5,000 base to crystallize significant ETH profits too early.
Variant C’s 30-minute clock did the same thing at a slower pace — 12 times.
Before deploying any rebalance configuration: disable or widen the time trigger significantly, or remove it entirely. A time trigger in a trending market converts a disciplined ratio strategy into a mechanical trend-fighter. If you must use a time trigger, set it no shorter than 24 hours for multi-week backtests. Test it on a divergence scenario first.
Overall Performance Score, Strengths and Limitations
Wrong Tool for This Market
Variant B avoided loss by doing almost nothing. Variants A and C confirmed what the market condition predicted: rebalancing in a sustained divergence is capital destruction on a schedule. The strategy worked as designed. The design was wrong for August 2025.
🧭 What this strategy does well
- Variant B's wide threshold correctly minimized unnecessary trading activity
- Fee rate at 0.1% kept per-swap costs negligible — not the problem here
- End-date USDT conversion provided clean P&L crystallization
- The A/B/C comparison cleanly demonstrates the threshold sensitivity lesson
🚫 What went wrong this month
- All variants dramatically underperformed passive HODL (est. −$418 behind on Variant B)
- Time triggers on Variants A and C turned a ratio strategy into a time-based one
- 50/50 static allocation offers no way to reduce exposure to the declining asset mid-session
- Strategy has no mechanism to detect or respond to fundamental asset divergence
Quick Takeaways
- Wide thresholds lose less in trending divergence — but still lose vs. HODL
- Time-based triggers dramatically increase trade frequency and damage in one-sided markets
- Rebalancing is a mean-reversion strategy — deploy it only when you expect mean reversion
- The best rebalance frequency in a strong trend is zero
- Always benchmark against HODL before deployment — August 2025 passive holding beat every active variant
How did passive HODL compare?
If you had simply bought $2,500 of BTC and $2,500 of ETH on July 18 at market open prices and held through September 15 without touching anything, here’s how it compares:
The opportunity cost of running the best-performing rebalance variant versus doing nothing: approximately −$418.68 in a single backtest window.
HODL didn’t require strategy, timing, parameter selection, or any active management. It simply held ETH through its 24% surge and absorbed BTC’s decline — the natural portfolio math worked in its favor precisely because no bot intervened to correct the “imbalance.”
Before you run this playbook, check these off.
Use this as your go/no-go checklist before deploying this exact parameter set.
🧠 Market Suitability Matrix
| Market Condition | Rating | Strategic Notes |
|---|---|---|
| Both assets sideways / choppy | ★★★★★ Excellent | Harvest spreads via consistent mean reversion. |
| One asset dips, then recovers | ★★★★★ Ideal | Buy dips, capture spreads during recovery. |
| Both assets in a mild bull market | ★★★★☆ Good | Trim winners to fund converging laggards. |
| One asset strongly outperforms | ★★★☆☆ Moderate | Systematically sells winner; missed Aug 2025 ETH surge entirely |
| Both assets are in steep decline | ★★☆☆☆ Risky | Redistributing losses without harvestable spreads. |
| One asset in the structural breakdown | ★☆☆☆☆ Poor | Buying declining assets; concentration risk increases. |
| Highly correlated assets | ★☆☆☆☆ Poor | Choose moderate correlation to capture spreads. |
How to tune this playbook for different scenarios.
Disclaimer: All data sourced from CryptoGates Rebalance Backtest Bot. Results are historical simulations using Binance 1-minute OHLCV data. Past backtest performance does not guarantee future live trading results. DYOR.
Battle-Test Your Strategy
Before the Market Does.
Eliminate guesswork with institutional-grade backtesting for DCA, Grid, and Rebalance bots. Real historical data. Real-world results.
