Skip to main content
rebates
arbitrage
hft
crypto trading
trading fees
backtest

50,000 Arb Bot Trades: No Rebates = Liquidation

Hard numbers from a 9-day, 50,388-trade backtest proving that in HFT and arbitrage environments, commission isn't an expense—it's the primary risk factor. And how a 40% rebate is the only mathematical fix.

Sliceback Team
7 min

The Setup

Test window: 9 days, 14 hours, 22 minutes. Total executions: 50,388 trades across BTC/USDT, ETH/USDT, and SOL/USDT perpetuals on Bybit and MEXC. Infrastructure: Co-located VPS (Frankfurt node), WebSocket order book streams at 10ms tick intervals, REST fallback for order placement under congestion.

Average gross alpha per trade: +0.052%. That number looks fine until you run the fee math. Then it looks like a suicide note.


Part I: The Mathematical Wall

How 0.1% Kills a 0.052% Edge

The arithmetic is not complex. It is just brutal.

Every arbitrage leg requires two fills: entry on Exchange A, exit on Exchange B. In almost all real-world arb scenarios, at least one of those fills—usually both—hits as a Taker order.

Here's why that's non-negotiable:

  • Arbitrage windows are measured in milliseconds. A limit order (Maker) sits in the order book and waits. Waiting is not a strategy—it's a donation to whoever closes the spread before you.
  • In our API logs, the average detected spread persistence on BTC/USDT cross-exchange was 140–180ms before mean-reversion. At 100ms execution latency, a Maker strategy catches roughly 30% of identified opportunities. At 40ms (our optimized pipeline), that number rises to ~71%—but even then, partial fills and slippage force Taker fallback on the aggressive leg.

Technical Note: We debugged the execution pipeline after noticing fill rate degradation during the London/NY session overlap. The incoming JSON on MEXC's REST endpoint revealed a 200ms processing lag during high-volatility windows—turning theoretical Maker fills into either missed trades or forced Taker market orders. WebSocket order placement (where supported) reduced this to 38ms average. The lesson: your fee tier is irrelevant if your execution model assumes latency you don't actually have.

The Fee Stack, Unvarnished

See also: Hidden Trading Costs: Broker Commissions and Rebates

Fee ComponentRateTrades Affected
MEXC Taker0.10% per side~94% of all legs
Bybit Taker0.10% per side~91% of all legs
Effective round-trip cost~0.20%Per completed arb cycle

With a gross alpha of +0.052% and a round-trip cost of ~0.20%, every single trade is net negative by approximately −0.148% before slippage.

Apply that across 50,388 trades on a hypothetical $50,000 starting capital:

Net PnL per trade ≈ −0.148%
Cumulative fee drag ≈ −0.148% × 50,388 = −74.57% of capital
Realized drawdown (accounting for partial winners): −59.3%

That's not a bad month. That's a structural liquidation trajectory baked into the model at inception.


Part II: The Rebate Solution

What a 40% Rebate Actually Does to Your P&L

A rebate—cashback on commissions paid—doesn't sound exciting. It also doesn't lose you 59% of your deposit. Here's the recalculation with a 40% rebate on Taker fees via a program like Sliceback:

Standard Taker fee:     0.10% per side
Post-rebate effective:  0.10% × (1 − 0.40) = 0.06% per side
Round-trip cost:        ~0.12% (vs. 0.20% baseline)

Now rerun the trade stack:

Net PnL per trade ≈ 0.052% − 0.12% = −0.068%

Still negative. But now the alpha gap is 0.068% instead of 0.148%. That's the difference between a strategy that's dead on arrival and one that reaches break-even with modest optimization to spread detection or execution latency.

Push execution latency from 100ms to 40ms (achievable with WebSocket + co-location), and fill rate on tighter spreads improves. Alpha per trade edges up to ~0.065–0.075%. At that point:

Net PnL per trade (optimized + rebate) ≈ +0.005% to +0.015%

Marginal. But positive. And at 50,000+ trades over 9 days, marginally positive is the entire business model. More details on how this works for automated systems: rebate for scalpers and trading robots.

Comparison Table: The Only Table That Matters

MetricNo Rebate40% Rebate (Sliceback)
Taker fee per side0.100%0.060%
Round-trip cost~0.200%~0.120%
Gross alpha per trade+0.052%+0.052%
Net PnL per trade−0.148%−0.068%
Cumulative drag (50,388 trades)−74.57%−34.26%
Realized drawdown−59.3%−21.4%*
Strategy viability❌ Guaranteed ruin⚠️ Breakeven threshold
With latency optimization (+0.02% alpha)❌ Still ruin✅ Net positive

*Drawdown figure accounts for gross winning trades partially offsetting fee drag. The rebate scenario retains enough capital to optimize and iterate.

Technical Note: The −21.4% drawdown figure in the rebate column is not a success story in isolation. It becomes one when you recognize that it represents a survivable loss curve—the strategy lives to be tuned. The −59.3% figure is not survivable at standard retail leverage. At 5x leverage, that's a margin call at approximately trade #18,000. The bot doesn't fail at the end of the test. It fails in the middle.


Part III: Insider Insight — Why Exchanges Love You and Won't Help You

The Liquidity Paradox

Here's the cynical truth about exchange fee structures: arbitrageurs are the most valuable participants in a fragmented market, and exchanges know it.

Cross-exchange arb bots do the dirty work of price synchronization. When BTC/USDT on Bybit diverges from MEXC by 0.15%, arb bots close that spread in under 200ms. This provides genuine price discovery and reduces spread for every other participant in the order book. Exchanges benefit directly—tighter spreads attract more flow, more flow means more fees collected.

And yet, those same exchanges will not give you VIP-tier fee discounts without institutional trading volumes. Bybit's VIP 1 tier (0.08% Taker) requires $1M+ in 30-day volume. VIP 4 (0.06%) requires $25M+. For a retail or semi-institutional arb operation running $50K–$500K capital, those thresholds are structurally unreachable.

The exchange wants your volume. It wants your liquidity provision. It will not price that service fairly until you're already large enough not to need the discount.

Where Rebate Programs Fill the Gap

Third-party rebate programs like Sliceback operate by aggregating volume across multiple sub-accounts and client flows to unlock institutional fee tiers at the broker level, then passing a portion of that discount back to the end user. The mechanics are not complicated:

  • Your trades route through an affiliated structure with pre-negotiated fee agreements.
  • The exchange pays a rebate to the aggregator based on total volume.
  • The aggregator returns 40%+ of that rebate to you.

This is not a loophole. It is the standard operating model for every prop desk and fund that isn't trading nine-figure monthly volumes. At Tier 1 fee levels, the alpha is non-existent. The entire edge in high-frequency arb is captured—or destroyed—in the 0.04% gap between what a retail account pays and what an institutional desk pays.

The 50,388-trade backtest doesn't lie. The math is not a warning. It's a proof.


Final Audit

The strategy described above—cross-exchange perpetual arbitrage on BTC, ETH, SOL—is not exotic. It is well-understood, well-competed, and razor-thin. What separates a sustainable operation from a deposit graveyard is not a better signal or a smarter entry logic.

It's 0.04%. Applied fifty thousand times.

Run your numbers before you run your bot. If you're new to the topic, start with the complete guide to scalping rebates.


Frequently Asked Questions

Does connecting to Sliceback change my trading conditions on the exchange? No. Your order book access, execution speed, and fee display on the exchange remain identical. The rebate is calculated separately on Sliceback's side based on your confirmed trading volume. Read more about the service's security: how safe are rebate services.

Can I use this with an automated trading bot? Yes. Rebates apply to all trades executed under your linked UID, regardless of whether they are placed manually or via API. Your bot's REST/WebSocket calls are tracked at the exchange level and reflected in your daily volume report.

Does a rebate affect the size of my trading fees? No. Your fees on the exchange remain standard (or even lower if you have VIP status). The rebate is paid out of the portion the exchange has already collected and passed to its partners.

Can I connect an existing account? No. Rebates apply only to accounts registered via our referral link. If you already have an account on the exchange, you will need to create a new one using the partner link to qualify.

Ready to start earning?

Create a free account and get cashback on every trade

Get Started