Author: CryptoPunk
Many crypto traders have experienced the same disappointment: strategies look stable and profitable in backtests, but once live, returns quickly shrink or even turn into losses. The problem is often not "misjudging the direction," but underestimating trading costs, especially slippage.
In faster-changing, more volatile, and more fragmented crypto markets, slippage is not a trivial decimal point but a real threshold that determines whether a strategy can survive. A deviation of 2 or 3 basis points can, in high-turnover strategies, wipe out all the paper alpha.
This article, based on long-term backtests of BTC/USDT and ETH/USDT, aims to answer a very practical question: To what extent does slippage erode strategy returns, and which strategies are most vulnerable to slippage?
1. Introduction: Why is slippage always underestimated?
Traders tend to underestimate slippage for three main reasons:
First, many backtests assume trades occur at close, open, or mid-prices, which is inherently optimistic. Second, many only account for fees, ignoring slippage, let alone the bid-ask spread at entry and exit. Third, many assume slippage is fixed, but in reality, it varies with volatility, volume, order size, and market liquidity.
This is why many strategies look good in Excel or backtest frameworks but fall apart in live trading. Profits are not as large as expected, and costs are much higher.
2. Methodology: Backtest design for BTC/ETH
This study keeps the current strategy and slippage framework unchanged, only extending the time range and output results.
- Assets: BTCUSDT, ETHUSDT
- Data: Binance Vision public spot 1-minute Kline
- Sample period: 2020-01-01 to 2025-12-31
- Run date: 2026-03-15
- Note: As of the actual check on 2026-03-15, Binance Vision's public spot 1m data returns 404 after January 2026, so this paper uses the latest available data up to 2025-12-31.
- Execution rule: signals generated at the close of the current bar, executed at the open of the next bar.
To facilitate reproducibility, the core execution parameters are as follows:
| Parameter | Setting |
| --- | --- |
| Initial capital | 100,000 USDT |
| Default fee rate | 0.05% per side (~5 bps) |
| Round-trip fee | approx. 10 bps, excluding slippage |
| Order mode | proportional to account equity |
| Default order size | 15% of account equity |
| Leverage | 1x |
| Allow both directions | Yes |
Strategies are categorized into three types:
- Low frequency: 20/50 MA trend following, 1H
- Medium frequency: RSI + MA filtering, 15min
- High frequency approximation: short-term mean reversion, 5min
Slippage models include:
- Fixed bps: 1 / 3 / 5 / 10 / 20 bps
- Volatility-related slippage
- Impact-based volume slippage
- Asymmetric bid-ask slippage
- Extreme market penalty
The core conclusion mainly relies on the "extreme_volume_impact + fees" scenario, as it more closely reflects real trading conditions with "volatility amplification + bilateral costs."
3. Backtest results: Focus on the key sets
Looking only at gross returns, many strategies still tell a story; but once fees and slippage are included, the story quickly ends.
A typical example is BTC high-frequency mean reversion:
- Costless net profit: 84,534
- Fees only: -99,168
- Fees + slippage: further worsened to -99,896
- Total trades: 36,008; fees paid: 66,456; slippage costs: 46,966
This shows the problem isn't just "slippage a bit high," but that the strategy's per-trade edge isn't enough to withstand costs; combined, they wipe out all gains.
On the other hand, ETH low-frequency trend strategy is among the few that remain profitable after costs:
- Costless net profit: 48,948
- Fees only: 23,664
- Fees + slippage: still positive at 13,463
This indicates slippage doesn't make all strategies worse, but it reveals which strategies have enough edge to survive costs and which only look profitable in backtests.
To visualize cost erosion more clearly, here is a summary table. The "fees + slippage" scenario uses the "extreme_volume_impact" model from this article.
| Asset | Strategy | Gross profit | Fees only | Fees + slippage | Fee cost | Slippage cost | Number of trades |
| --- | --- | --- | --- | --- | --- | --- | --- |
| BTC | Low-frequency trend | 10,557 | -8,617 | -14,898 | 19,009 | 7,118 | 1,268 |
| BTC | Medium-frequency RSI+MA | 169 | 94 | 60 | 75 | 35 | 5 |
| BTC | High-frequency mean reversion | 84,534 | -99,168 | -99,896 | 66,456 | 46,966 | 36,008 |
| ETH | Low-frequency trend | 48,948 | 23,664 | 13,463 | 22,322 | 10,238 | 1,238 |
| ETH | Medium-frequency RSI+MA | 5 | -175 | -260 | 180 | 84 | 12 |
| ETH | High-frequency mean reversion | -29,338 | -99,665 | -99,934 | 39,020 | 60,551 | 31,421 |
This chart compares net profits under different slippage models. Fixed bps are just the starting point; once slippage interacts with volatility, volume impact, and extreme conditions, strategy returns decline sharply. For high-frequency strategies, switching from "fixed slippage" to "dynamic slippage" often results in profits disappearing altogether.
The comparison across models shows that fixed bps is the most conservative estimate; once market conditions cause slippage to fluctuate with volatility and volume, many strategies that barely break even in backtests will fall into losses.
4. How slippage erodes returns
The most frightening aspect of slippage isn't just "reducing some profit," but often pushing strategies from profit territory into losses.
In this study, 54 cases were identified where gross profit was positive but net profit was negative; within the model comparison alone, there are 40 such cases.
Typical failure examples include:
- BTC low-frequency trend: gross profit 10,557, fees-only loss -8,617, slippage loss -14,898
- ETH medium-frequency RSI+MA: gross profit only 4.53, fees turn it negative, and slippage worsens the loss
- BTC high-frequency mean reversion: profitable on paper, but costs nearly wipe out all gains
This explains why "backtest profits but live losses" are so common in crypto markets. Many strategies are not wrong in logic but are built on the false assumption that trading costs are negligible.
The above chart shows the net value of the BTC high-frequency mean reversion strategy. The blue line is the backtest net value ignoring costs; the green line includes fees and slippage. The former appears as a continuously compounding curve, while the latter is almost flattened to near zero.
Cost structure analysis further clarifies the issue. Using the reference slippage model:
- BTC high-frequency strategy's slippage cost is about 347% of gross profit
- Fee cost is about 491% of gross profit
- In low-frequency trend strategies, slippage costs are about 63% of gross profit
- In ETH low-frequency trend strategies, slippage costs are about 22% of gross profit
This means low-frequency strategies are more "costs squeezed," while high-frequency ones are "profits directly swallowed."
When considering returns, Sharpe ratio, and maximum drawdown, the impact of costs becomes even clearer:
| Asset | Strategy | Scenario | Net profit | Sharpe | Max drawdown |
| --- | --- | --- | --- | --- | --- |
| BTC | Low-frequency trend | No costs | 10,557 | 0.23 | -13.99% |
| BTC | Low-frequency trend | Fees + slippage | -14,898 | -0.25 | -24.32% |
| BTC | High-frequency mean reversion | No costs | 84,534 | 1.22 | -7.33% |
| BTC | High-frequency mean reversion | Fees + slippage | -99,896 | -13.10 | -99.90% |
| ETH | Low-frequency trend | No costs | 48,948 | 0.62 | -22.08% |
| ETH | Low-frequency trend | Fees + slippage | 13,463 | 0.24 | -25.22% |
| ETH | High-frequency mean reversion | No costs | -29,338 | -0.47 | -36.72% |
| ETH | High-frequency mean reversion | Fees + slippage | -99,934 | -11.35 | -99.93% |
5. Why high-frequency strategies are most vulnerable to slippage
High-frequency strategies are most easily killed by slippage, not necessarily because their directional judgment is wrong, but because their profit margins are extremely thin.
Common traits of high-frequency strategies:
- Very small profit per trade
- Extremely high trading frequency
- Highly sensitive to execution prices
In this backtest, the average cumulative slippage costs under the reference model are:
- High-frequency: 53,758
- Low-frequency: 8,678
- Medium-frequency: 59
This indicates that slippage's main impact is concentrated on high-turnover strategies.
From a trading frequency perspective, the average profile under the reference model is:
| Frequency | Avg net profit | Avg cumulative slippage | Avg realized slippage | Avg number of trades |
| --- | --- | --- | --- | --- |
| High | -99,915 | 53,758 | 5.65 bps | 33,714 |
| Low | -718 | 8,678 | 2.08 bps | 1,253 |
| Medium | -100 | 59 | 2.32 bps | 9 |
This chart shows that the "net profit erosion" is sharply higher for high-frequency strategies, confirming that slippage impact is heavily skewed toward high-turnover approaches. Many high-frequency systems don't necessarily fail to make money but can't earn enough to offset the continuous friction.
More importantly, slippage and trading frequency are not simply linear; they accelerate under high volatility and large order sizes.
For example, in the reference model, the average realized slippage for high-frequency strategies under high volatility is:
- BTC: 2.33x the low-volatility level
- ETH: 3.99x
When order sizes increase, this erosion becomes even more pronounced:
- BTC high-frequency realized slippage rises from 2.24 bps to 5.70 bps
- ETH high-frequency realized slippage rises from 3.40 bps to 16.34 bps
This is illustrated in the next chart, which shows cumulative slippage loss at different order sizes. The curve is convex, not linear, meaning that increasing position size causes slippage costs to grow disproportionately. For ETH high-frequency strategies, increasing position from 5% to 35% can cause slippage to worsen rapidly.
This is a crucial insight many traders overlook: position scaling is not just a simple multiplication; slippage tends to grow convexly, so strategies that work at small scale may become unprofitable when scaled up.
6. BTC vs ETH: differences in slippage
Many traders assume BTC is "more expensive," so slippage should be higher. But the actual backtest results tell a more nuanced story.
Total slippage costs under the reference model:
- BTC average cumulative slippage: 18,039
- ETH average cumulative slippage: 23,624
In terms of realized slippage per unit traded (bps):
- BTC: 2.57 bps
- ETH: 4.13 bps
Breaking down by strategy:
- High-frequency: BTC 3.53 bps vs ETH 7.76 bps
- Low-frequency: BTC 1.87 bps vs ETH 2.29 bps
- Medium-frequency: BTC 2.31 bps vs ETH 2.34 bps
Putting BTC and ETH side by side makes it clearer: despite ETH's higher total slippage, its per-trade realized slippage is consistently higher across strategies, indicating ETH's liquidity friction is more sensitive to trading costs.
This suggests that, in the long run, ETH's liquidity costs are more impactful, especially in high-frequency and high-volatility scenarios.
7. Conclusion: Slippage is not a small error but a critical threshold
This study's key takeaways are clear:
First, slippage is not just a minor parameter in backtests; it is a crucial factor that determines whether a strategy can be practically traded. Second, many strategies that look profitable in backtests fail in live trading because they assume near-zero trading costs. Third, high-frequency strategies are most vulnerable to slippage because they rely on tiny per-trade margins and high turnover. Fourth, ETH's unit slippage pressure is generally higher than BTC's, especially during volatile and high-turnover periods. Fifth, larger order sizes cause slippage costs to grow convexly, not linearly.
For crypto traders, the real questions are not "How much can this strategy make in backtests?" but:
- After bilateral fees, how much remains?
- After realistic slippage, how much remains?
- Can it survive during high-volatility, low-liquidity periods?
- Will it turn from profitable to losing when scaled up?
Without answers to these, high backtest returns are likely hiding critical cost assumptions.
