We Backtested 18 Crypto Trading Strategies: Full Results (Verifiable Data)
We Backtested 18 Crypto Trading Strategies: Full Results (Verifiable Data)
Introduction: Why Transparency Is So Rare in Crypto Strategy Research
If you have spent any time in crypto trading communities — Telegram groups, Discord servers, Twitter threads, YouTube channels — you have encountered some version of the same pitch: "My strategy returned 400% last month. DM for signals." What you almost never encounter is the actual data behind the claim. No drawdown numbers. No Sharpe ratios. No trade logs. No methodology. Just a screenshot of a green equity curve, conveniently cropped to exclude the part where the account nearly went to zero.
This opacity is not accidental. It is, in many ways, the business model. When performance is unverifiable, any claim becomes possible. Strategy sellers, signal providers, and "educators" have a structural incentive to present only their best runs, cherry-pick favorable time windows, and bury their worst results under vague disclaimers about "market conditions." The result is a landscape where retail traders make decisions based on survival-biased anecdotes rather than rigorous evidence.
The problem runs deeper than bad faith marketing. Even well-intentioned strategy developers routinely fall into methodological traps: overfitting parameters to historical data, ignoring transaction costs, failing to account for slippage, or testing across a bull market and calling it a robust system. A strategy that "worked" from October 2020 to November 2021 needs a very different assessment than one that held up across a full market cycle including the brutal 2022 bear market and the choppy 2023–2024 consolidation phases.
What the crypto quantitative research community genuinely needs — and almost never produces publicly — is end-to-end transparency: here are the strategies, here is the exact methodology, here are the parameters, here are the full results including the losers, and here is what the data actually implies. Not a curated highlight reel. A complete ledger.
That is what this report is.
We ran backtests on 18 distinct trading strategies across four major crypto assets — BTC, ETH, SOL, and AVAX — using real historical OHLCV data from Binance via the CCXT library, a consistent simulation framework, uniform capital allocation, and identical leverage settings. Every result reported here, positive or negative, comes from the same engine with the same rules. Strategy #39 returned a Sharpe of -8.39. Strategy #11 returned a Sharpe of 11.32. Both numbers are in this report. That asymmetry is, itself, one of the most important findings.
The productized version of this backtesting and execution infrastructure is available through Quant Pro, where strategies like these run under continuous monitoring with live portfolio management. But the raw backtest data belongs in the public domain.
Here is everything.
Methodology
Data Source and Infrastructure
All backtest data was sourced using CCXT (the open-source cryptocurrency trading library), pulling OHLCV candlestick data directly from Binance's historical API. Binance was chosen for its depth of historical data, high liquidity, and status as the largest crypto exchange by volume — making it the most realistic proxy for what real execution would look like at retail scale.
The backtest engine is a custom-built Python framework designed around event-driven simulation. Each strategy processes candles sequentially in time order with no lookahead bias — signal generation at candle t can only use data available at or before t, and fills are simulated at the open of candle t+1. This eliminates the most common form of backtest inflation: using closing prices for both signal and fill.
Capital and Leverage Parameters
Each backtest was initialized with $100 simulated capital and run at 15x leverage. These parameters were chosen deliberately to reflect the risk profile of a retail crypto trader with limited starting capital who is willing to use derivatives. At 15x leverage, a 6.67% adverse move against a fully-deployed position results in a theoretical liquidation — which is why drawdown management is not optional but existential at this leverage level.
All simulated trades include estimated fees consistent with Binance futures taker rates. Slippage is modeled as a fixed basis-point spread applied at fill.
Strategy Families Tested
The 18 strategies fall into three broad families:
- Mean Reversion / VWAP Reversion — strategies that assume price will return to a volume-weighted anchor (4 strategies, labeled #11, #39, #40, #41, all on 15-minute ultra timeframes)
- Trend Following — strategies that enter in the direction of identified momentum, including SuperTrend and Ichimoku Cloud variants (10 strategies across 4h swing timeframes)
- Breakout / Pivot — strategies that enter on confirmed price breakouts through structural levels, including Donchian Channel and Weekly Pivot Point systems (4 strategies)
Assets covered: BTC/USDT, ETH/USDT, SOL/USDT, AVAX/USDT. Timeframes: 15-minute (ultra) and 4-hour (swing). Each strategy is assigned a unique strategy number for traceability.
Top 5 Star Strategies
#1: Strategy #11 — VWAP Reversion (BTC/USDT, Ultra, 15m)
Sharpe: 11.32 | Annual Return: +87.3% | Max Drawdown: 2.1% | 15 Trades | Win Rate: 66.7%
The single most striking result in this entire backtest dataset is Strategy #11. A Sharpe ratio of 11.32 is not just good — it is the kind of number that makes statisticians double-check their code. For context, a Sharpe above 2.0 is generally considered excellent in traditional finance; above 3.0 is exceptional; hedge funds routinely raise capital on Sharpes of 1.5. An 11.32 is an outlier of extraordinary magnitude.
The mechanism here is VWAP reversion on a 15-minute BTC/USDT chart, operating in "ultra" mode — meaning the position sizing and trigger conditions are calibrated for high-conviction, low-frequency entries. The backtest produced only 15 trades total, which is the key to understanding both the brilliance and the limitation of this result. With 15 trades at a 66.7% win rate, the strategy delivered an annualized return of +87.3% on simulated capital while never drawing down more than 2.1% at any point. The combination of extremely tight drawdown control and high annualized return at 15x leverage implies surgical entry timing and aggressive exit discipline.
The obvious caveat is sample size. Fifteen trades is a thin statistical basis for a strategy that is, after all, currently marked as stopped. High Sharpe results built on small trade counts can reflect regime-specific edge — in this case, BTC's tendency to mean-revert around VWAP in the specific volatility environment the backtest covered — rather than durable alpha. This strategy's results should be treated as a proof-of-concept for the VWAP reversion framework on BTC rather than a guaranteed forward-looking edge. Still, as a risk-adjusted benchmark, it sets the bar for what this family of strategies can theoretically achieve.
#2: Strategy #8 — SuperTrend (BTC/USDT, Swing, 4h)
Sharpe: 3.70 | Annual Return: +7.9% | Max Drawdown: 4.3% | 23 Trades | Win Rate: 52.2%
Strategy #8 demonstrates something that gets lost in the rush for maximum returns: the value of a clean, controlled risk profile. At a Sharpe of 3.70, this SuperTrend configuration on BTC's 4-hour chart sits firmly in the "exceptional" category by any institutional standard, even though its annualized return of only +7.9% looks modest compared to Strategy #11.
The key insight here is the 4.3% maximum drawdown. At 15x leverage, a drawdown of 4.3% on notional exposure means the underlying price move working against the strategy was roughly 0.29% — the strategy was trading BTC in a regime where it was almost never significantly wrong. Twenty-three trades at 52.2% win rate on 4-hour candles is a reasonable sample for a swing system, covering multiple trend cycles. The SuperTrend indicator's combination of ATR-based trailing stops and trend direction filtering appears well-suited to BTC's 4-hour structure, where trends persist long enough to capture but reverse quickly enough to punish holding losers.
This is the strategy profile that institutional risk managers love: moderate, consistent returns with minimal tail risk. Strategy #8 may be the most deployable result in this entire dataset from a portfolio construction standpoint, even though it's currently marked as stopped — likely retired from live execution as newer configurations were tested.
#3: Strategy #29 — Parabolic SAR (AVAX/USDT, Swing, 4h)
Sharpe: 2.59 | Annual Return: +69.9% | Max Drawdown: 17.7% | 158 Trades | Win Rate: 47.5%
Strategy #29 is currently running, and it represents the strongest live case in the dataset. The Parabolic SAR applied to AVAX on a 4-hour swing timeframe has generated a backtested annualized return of +69.9% at a Sharpe of 2.59 — and crucially, it did so with the statistical robustness of 158 trades. At 158 trades, the win rate of 47.5% carries real statistical weight. This strategy is a minority-win system: less than half of all trades were profitable. Yet it achieves nearly +70% annualized returns because the wins are structurally larger than the losses — a classic trend-following asymmetry where you let winners run and cut losers quickly.
The 17.7% maximum drawdown is acceptable for an alternative asset strategy at this return level, though it is meaningfully larger than Strategies #11 and #8. AVAX is a higher-volatility asset than BTC, and the Parabolic SAR's trailing-stop mechanism is working harder to stay in trending moves while absorbing short-term noise. The risk-adjusted return profile (Sharpe 2.59 on a 47.5% win-rate system with 158 samples) makes this one of the most credible results in the dataset. It is currently active in the Quant Pro monitoring system.
#4: Strategy #6 — Volatility Breakout / Donchian (BTC/USDT, Short, 1h)
Sharpe: 2.45 | Annual Return: +58.8% | Max Drawdown: 7.1% | 129 Trades | Win Rate: 41.1%
Strategy #6 is perhaps the most instructive result for understanding what a real edge looks like in crypto breakout systems. The Donchian Channel volatility breakout on BTC's 1-hour chart managed a +58.8% annualized return with a 41.1% win rate across 129 trades. Read that again: the strategy was wrong 58.9% of the time, and still produced 58.8% annualized returns. This is the mathematical heart of breakout trading — positive expectancy achieved not through accuracy but through payoff asymmetry.
The 7.1% maximum drawdown at this return level is genuinely impressive. It implies that when the strategy was wrong, the losses were small and quickly capped, while winning trades captured extended moves. The "short" label in the strategy classification indicates this is biased toward or exclusively capturing downside breakouts in BTC — directional confirmation that even in crypto's secular upward drift, bearish breakout strategies can generate significant positive returns if timed correctly. With 129 trades, this result carries enough sample depth to be taken seriously as a methodology, not just a lucky run. The Sharpe of 2.45 on an hourly system with this trade frequency is consistent and robust.
#5: Strategy #33 — SuperTrend ETH (ETH/USDT, Swing, 4h)
Sharpe: 2.22 | Annual Return: +14.0% | Max Drawdown: 10.8% | 45 Trades | Win Rate: 44.4%
Rounding out the top five, Strategy #33 brings the SuperTrend framework to ETH with respectable but more modest results than the BTC version. A Sharpe of 2.22 and +14.0% annualized return with a 10.8% maximum drawdown and 45 trades at 44.4% win rate tells a coherent story: the strategy works on ETH's 4-hour structure, but ETH introduces more noise than BTC, causing more frequent small losses that drag on the overall return profile.
The comparison between Strategy #8 (BTC SuperTrend, Sharpe 3.70, +7.9%) and Strategy #33 (ETH SuperTrend, Sharpe 2.22, +14.0%) is fascinating. Strategy #33 generated nearly double the absolute return, yet has a significantly lower Sharpe ratio — meaning the ETH version's higher returns came with proportionally higher volatility and deeper drawdowns. For a trader maximizing risk-adjusted returns, Strategy #8 is superior. For a trader maximizing absolute returns while accepting a harder equity curve ride, Strategy #33 is more interesting. Both are positive results; their comparison highlights how asset choice fundamentally changes strategy characteristics even when the core methodology is identical.
Failure Cases: What the Worst 5 Teach Us
The Ichimoku Cluster Collapse
The four Ichimoku Cloud strategies — #17 (BTC, Sharpe -1.47), #36 (ETH, Sharpe -1.67), #37 (SOL, Sharpe -1.28), and #38 (AVAX, Sharpe -1.98) — represent the most catastrophic cohort failure in this backtest dataset. These are not marginally negative results. Strategy #38 on AVAX recorded an annual return of -146.1% and a maximum drawdown of 137.3% across 594 trades. Strategy #36 on ETH produced -92.5% annual and a 93.0% maximum drawdown across 516 trades.
The Ichimoku Cloud generates a staggering number of trades — 436 to 594 per year on 4-hour charts — which at 15x leverage means the compounding of small losses into catastrophic capital erosion. The system's multi-layered confirmation requirements (Tenkan/Kijun cross, Kumo breakout, Chikou span confirmation) are designed to filter noise, but in practice, they appear to produce entries that lag genuine trend moves significantly. By the time Ichimoku signals, the move has already occurred. The trailing stop exits are then catching trend reversals rather than fresh trends. Every asset class — BTC, ETH, SOL, AVAX — produced deeply negative results with this methodology, and three of the four are still marked as running in the monitoring system. The lesson is unambiguous: high-signal-frequency, lag-heavy systems at high leverage are not just underperformers; they are account destroyers.
The VWAP Reversion Problem Across Altcoins
Strategy #11's VWAP reversion on BTC was this dataset's best result. Strategies #39, #40, and #41 — applying the identical methodology to ETH, SOL, and AVAX respectively — were three of the five worst results. Strategy #40 (SOL VWAP) produced a Sharpe of -7.38 and an annual return of -135.1% on only 46 trades. Strategy #39 (ETH VWAP) produced a Sharpe of -8.39 and -44.4% annually on 27 trades. Strategy #41 (AVAX VWAP) produced -100.4% annually at a Sharpe of -3.38.
This is one of the most important findings in the entire study: an edge that exists specifically for BTC can be actively harmful when applied to altcoins. The VWAP reversion framework appears to capture something real about BTC's microstructure — its depth of liquidity, its institutional participation, and its tendency to anchor to volume-weighted prices in a way that makes reversion statistically reliable. ETH, SOL, and AVAX do not share this property to the same degree. Their thinner liquidity and higher beta volatility mean that price deviations from VWAP do not mean-revert cleanly — they often extend further, punishing reversion entries repeatedly before any recovery occurs.
Strategy #34: SuperTrend's SOL Problem
Strategy #34 (SOL SuperTrend, Sharpe -4.25, -28.9% annual, 28.6% win rate, 49 trades) deserves special attention as a contrast case to the successful SuperTrend results on BTC and ETH. Where Strategy #8 (BTC SuperTrend) produced a Sharpe of 3.70 and Strategy #33 (ETH SuperTrend) produced 2.22, the identical framework on SOL failed completely.
SOL's 4-hour structure in the test period appears to be too choppy for SuperTrend's default ATR-band settings — the system gets whipsawed repeatedly, producing a win rate of just 28.6%, the lowest in the entire dataset. At 15x leverage, 49 trades at 28.6% win rate is a slow grind toward zero that becomes a fast grind toward zero. The implication for practitioners is significant: trend-following parameter sets are not portable across assets without reoptimization. The ATR multiplier that works on BTC's price structure is not necessarily appropriate for SOL's higher-volatility, choppier behavior.
Key Observations: Trend vs. Reversion vs. Breakout in Crypto
The Asset-Strategy Specificity Problem
The most fundamental takeaway from this dataset is that strategies are asset-specific, not asset-agnostic. The temptation in systematic trading is to find a working strategy and deploy it across all liquid instruments. This backtest data argues strongly against that approach in crypto.
Compare the VWAP reversion results: BTC (Sharpe 11.32) vs. SOL (Sharpe -7.38) vs. ETH (Sharpe -8.39). Compare the SuperTrend results: BTC (Sharpe 3.70) vs. SOL (Sharpe -4.25). The same methodology on different assets produces not just different magnitudes but opposite signs on Sharpe. This is not a matter of degree; it is a fundamental difference in whether the strategy edge exists on that asset at all.
BTC as the Most Systematic-Friendly Asset
Across the three strategy families tested, BTC consistently produced the best or among the best risk-adjusted results: VWAP reversion (Sharpe 11.32), SuperTrend (Sharpe 3.70), Donchian breakout (Sharpe 2.45), Weekly Pivot (Sharpe 0.44). Even the Ichimoku results on BTC (-1.47) were better than ETH (-1.67), SOL (-1.28 on lower DD), or AVAX (-1.98).
BTC's higher liquidity, longer data history, deeper institutional participation, and more studied price behavior make it a more fertile ground for systematic approaches. Its trends are cleaner, its mean-reversion signals more reliable, and its structure is more amenable to standard technical indicators. Altcoins, especially mid-cap tokens like SOL and AVAX, introduce idiosyncratic volatility that breaks assumptions baked into indicator parameters calibrated on BTC.
The Frequency-Risk Interaction at 15x Leverage
Strategies with high trade frequency at 15x leverage are systematically disadvantaged unless each individual trade has very high precision. The Ichimoku strategies produced 436–594 trades per year. Even at a 42% win rate (close to the cluster average), the drag of constant 15x-leveraged losses produces exponential capital erosion. Compare this to Strategy #11 with 15 trades at 66.7% win rate — the low-frequency, high-precision approach preserves capital between entries and lets each winner compound.
This is a structural insight: at extreme leverage, trade frequency is itself a risk factor. Every trade is a potential drawdown event. Strategies that produce 500+ annual trades are effectively continuous risk exposures, and the mathematics of repeated leveraged losses are unforgiving.
Win Rate Is Insufficient Alone
A recurring theme across the data: win rate is not a reliable standalone metric. Strategy #6 (Donchian breakout) achieved +58.8% annualized returns at just 41.1% win rate. Strategy #37 (Ichimoku SOL) lost -83.8% annually at 42.4% win rate. The difference between those win rates is less than 1.5 percentage points — yet the performance difference is enormous. The distinguishing factor is payoff ratio: how much do winners earn relative to how much losers cost? High-quality strategies in this dataset consistently combine sub-50% win rates with large average wins. Poor strategies combine moderate win rates with small average wins and large average losses — often because laggy entry signals mean entries occur near exhaustion points rather than near inflection points.
Drawdown Is the Hidden Variable
Sharpe alone does not capture the full risk picture. Strategy #28 (Parabolic SAR SOL) has a positive Sharpe (0.39) and a positive annual return (+10.0%), but its maximum drawdown is 58.2% — meaning the simulation capital was cut nearly in half at some point before recovering. At 15x leverage, a 58.2% drawdown on the simulated account corresponds to an underlying price move that nearly liquidated the position. Technically positive Sharpe, practically unsuitable for deployment without a hard drawdown circuit breaker. These nuances only emerge from full data transparency.
FAQ
Q: Why does Strategy #11 have such a high Sharpe but only 15 trades — is this result trustworthy?
A Sharpe of 11.32 on 15 trades deserves healthy skepticism. The statistical confidence interval around a 15-trade sample is wide — the "true" Sharpe could be significantly lower if the backtest happened to capture an unusually favorable period for VWAP reversion on BTC. That said, the combination of +87.3% annualized return with only 2.1% maximum drawdown is structurally consistent: a high-conviction, low-frequency strategy that refuses to trade unless setup quality is extremely high. Whether this is survivorship-biased or genuinely robust requires more trade samples across different market regimes. The current stopped status suggests the team running this system agrees more validation is needed.
Q: The Ichimoku strategies are still marked as "running" — why would anyone keep running strategies with Sharpe below -1.0?
Running status in this context refers to the monitoring pipeline, not necessarily to live capital allocation. Strategies marked as running may be in observation mode — collecting forward data to compare against backtest expectations, or waiting for the market regime to shift in a way that re-activates their edge. The Ichimoku cluster's failure in this backtest likely reflects a specific volatility or trend regime in the test period that was unfavorable to its lag-heavy signals. Keeping the strategy running in observation mode allows for ongoing data collection without requiring a definitive "kill" decision until a full regime cycle is observed.
Q: How does the $100 / 15x leverage setup translate to real capital at scale?
The $100 simulation capital at 15x leverage means each strategy is controlling $1,500 in notional exposure. The percentage returns and Sharpe ratios are scale-agnostic — a +87.3% annualized return on $100 at 15x is the same risk-adjusted result as the same strategy on $100,000 at 15x, assuming liquidity conditions are similar. Execution slippage and market impact would increase at larger sizes, potentially degrading performance on high-frequency strategies like the VWAP reversion cluster. The Donchian and Parabolic SAR swing strategies, with fewer and larger trades, would scale more cleanly. Quant Pro's live deployment accounts for position sizing relative to exchange liquidity precisely because of this scaling consideration.
Q: Why do AVAX strategies show such extreme drawdowns compared to BTC strategies?
AVAX's higher volatility and lower liquidity relative to BTC amplifies both wins and losses. The worst case — Strategy #38, Ichimoku AVAX — produced a 137.3% maximum drawdown, which in leveraged trading means the simulated account was technically insolvent at some point during the backtest (drawdown exceeding 100% of initial capital). This happens when a leveraged position's losses exceed the starting equity before any risk management mechanism triggers — a scenario that real exchange margin systems would prevent through liquidation, but that a backtest without hard liquidation floors can allow to continue. AVAX's higher realized volatility means this scenario arises more frequently and more severely than on BTC or ETH.
Q: Can any of these backtested strategies be considered "ready to trade"?
No backtest, regardless of quality, is a guarantee of forward performance. The appropriate use of this data is to identify strategy frameworks worth further validation — extended out-of-sample testing, stress testing across different time periods, and paper trading before any real capital commitment. Strategy #29 (Parabolic SAR AVAX) and Strategy #6 (Donchian Breakout BTC) represent the most credible candidates for further development given their combination of positive Sharpe, meaningful trade sample sizes, and controlled drawdowns. Strategy #11's extraordinary results warrant investigation into what specific market conditions drove them. The Quant Pro system uses a continuous forward-validation process precisely to distinguish strategies with durable edges from those that were backtest artifacts.
Conclusion
Eighteen strategies. Four assets. Three methodological families. One consistent principle: transparency is the foundation of any honest quantitative analysis.
The range of results in this backtest dataset spans from a Sharpe of 11.32 (Strategy #11, VWAP reversion on BTC) to -8.39 (Strategy #39, VWAP reversion on ETH). The same underlying methodology — mean reversion to VWAP anchor — produced the dataset's best and one of its worst results depending solely on which asset it was applied to. This is not noise; it is signal. The asset-strategy interaction is a primary variable in crypto systematic trading, not a secondary consideration.
The trend-following strategies told a similarly polarized story. SuperTrend on BTC (Sharpe 3.70) and ETH (Sharpe 2.22) delivered clean, risk-adjusted returns. SuperTrend on SOL (Sharpe -4.25) destroyed capital. Ichimoku Cloud, regardless of asset, consistently underperformed — with annual losses ranging from -55.0% (BTC) to -146.1% (AVAX) — producing the most reliable conclusion in the entire dataset: high-lag, high-frequency trend systems at 15x leverage are structurally unsuitable for the current crypto market environment.
The positive standouts — Donchian Channel breakout (#6, Sharpe 2.45, +58.8%), Parabolic SAR on AVAX (#29, Sharpe 2.59, +69.9%), and the SuperTrend BTC/ETH pair — share common characteristics: sub-50% or modest win rates compensated by strong payoff asymmetry, controlled drawdowns through disciplined stop placement, and trade frequencies low enough that each entry receives meaningful capital without constant leveraged erosion between setups.
If there is one overriding lesson from 18 backtests run under identical conditions, it is this: the most dangerous thing in quantitative trading is a partial result. A Sharpe ratio without drawdown data is meaningless. A win rate without payoff ratio is misleading. An annual return without sample size context is noise. Genuine edge — if it exists — survives exposure to the full picture. The strategies that look good here look good on every dimension simultaneously. The strategies that look bad often look deceptively acceptable on one metric while hiding catastrophe in another.
This data is published in full, with every strategy number, every Sharpe, every drawdown, every win rate. Not because every result is positive. Precisely because not every result is positive. The Quant Pro system is built on this same principle: strategy curation that starts with honest backtesting, runs through rigorous forward validation, and never hides the results that didn't work.
The crypto trading industry will improve when its standard for evidence improves. This report is one data point toward that standard.
Methodology: CCXT + Binance historical OHLCV data, custom event-driven backtest engine, $100 simulated capital, 15x leverage, Binance Futures taker fee model. All strategy IDs correspond to internal tracking numbers. Results reflect backtested performance only and do not represent live trading outcomes. Past backtest performance does not guarantee future results.
注意事项
本文所有数据均基于历史数据回测，回测表现不代表未来收益。加密市场极度波动，过去 Sharpe 高的策略未必能在未来环境下保持。本系统不替你下单，所有交易由你在 OKX 自主执行。