The Bot Lab

We test the crazy ideas so you don't have to.

Virtual trading experiments. Real market data. Fully transparent rules. No real money. Every strategy is open — you see exactly how it works.

Experiments

Running

$10k

Virtual each

+2104.2%

Best bot

Real code, real data

Every bot is a real Python script running real backtests on real market data. No fake numbers, no cherry-picked charts.

Fully transparent

You see the exact rules, parameters, and tech stack. Every trade is logged and timestamped. No black boxes.

Graduation path

Bots that prove themselves over 3+ months get promoted to live trading with real money. From lab to reality.

How we backtest

Why these numbers actually mean something

Most backtests online are garbage. Cherry-picked timeframes, missing fees, curve-fit parameters, no validation. Here's what we do differently.

6 years of real data

Every backtest runs on 2,296 days of actual BTC/USD prices from CryptoCompare — Jan 2020 to today. That includes the 2020 COVID crash, the 2021 bull run, the 2022 bear market (-77%), and the 2024 rally.

Real trading fees included

0.1% per trade (Bybit spot taker fee). Every buy. Every sell. Applied in the order they happen. This is why our Scalper bot has a 65% win rate and still loses money — fees eat the edge.

Multiple market cycles

A strategy that only works in a bull market isn't a strategy. Our 6-year window forces every bot through both euphoria and despair. Strategies that collapse in bear markets get exposed.

Day-by-day equity tracking

We don't just report the final number. Every card shows the real portfolio value on every single day — including drawdowns, flat cash periods, and recoveries. No smoothing, no interpolation.

Parameter robustness tests

We don't just run one parameter set. We test 5-10 variants per bot. If a strategy only works with one specific setting (fragile), we label it. Only strategies that work across many settings (robust) earn trust.

We publish our failures

When our "gold nugget" Volume Spike bot went from +45% to -27% after extending the data, we wrote an article about it. Overfitting is the silent killer of backtests. We name it.

Typical crypto backtest vs. BotLab backtest

❌ What you usually see

• Cherry-picked 6-month bull market window
• No trading fees included
• One parameter set that "just happens to work"
• Final number only, no equity curve
• Only winners shown, losers hidden
• No code, no data source, no validation
• "Trust me bro"

✅ What we do

• 6 years of data across all market conditions
• 0.1% fee per trade, compounded correctly
• 5-10 parameter variants tested for robustness
• Day-by-day equity curve visible on every card
• All 23 bots shown, including the failures
• Open source Python framework, public data
• "Here's the code. Run it yourself."

Read the full methodology article

Still not perfect: 6 years of data captures one major Bitcoin cycle. True statistical significance would require decades. We are transparent about this. That's also why bots need to prove themselves live for 3+ months before graduating to real money.

BotLab Leaderboard

Live rankings

#	Bot	Return	Win Rate	Trades	Max DD	Robust
🥇	⏳ The Halving Bot Buy 6 months before halving, sell 12 months after. +91%.	+2104.2%	100%	2(~1/3yr)	0%	100%
🥈	📆 The Seasonality Bot Sell in May? We tested crypto seasonality.	+1482.8%	71.4%	7(~1/1yr)	0%	100%
🥉	🔄 The Contrarian Does the OPPOSITE of what RSI says. And it works.	+453.9%	55.6%	9(~1/1yr)	0%	100%
4	😱 The Panic Buyer Buys when everyone screams SELL	+391%	50%	4(~1/2yr)	-28%	60%
5	🌕 The Lunatic Full moon buy, new moon sell. Yes, really.	+214.8%	51.3%	78(~1/mo)	0%	—
6	🌀 The Fibonacci Bot Sacred geometry or trader astrology? We tested it.	+155.5%	80%	15(~1/5mo)	0%	—
7	🏃 The Momentum Bot Rides winning streaks, exits on first red day	+130%	35.9%	128(~1/mo)	0%	83%
8	🧬 The Entropy Bot Uses information theory to detect regime changes. Beats HODL.	+117.5%	60.9%	23(~1/3mo)	0%	100%
9	🐢 The Turtle Classic Donchian channel breakout — the original turtle trading	+102.2%	47.4%	95(~1/mo)	0%	100%
10	📉 The RSI Bot The most popular indicator. Does it actually work?	+46.5%	66.7%	9(~1/1yr)	0%	—
11	📅 The Weekender Monday buy, Friday sell. Every week.	+30.1%	51.5%	328(~1/w)	0%	—
12	⚡ The Scalper Buys every dip, sells every bounce. Speed kills.	+30%	68.9%	177(~1/mo)	0%	—
13	🐋 The Whale Watcher Follows the smart money on-chain	+28.8%	60%	5(~1/6mo)	0%	100%
14	📈 The Funding Rate Bot Trades against the crowd when leverage gets extreme	+27.9%	50%	2(~1/1yr)	0%	—
15	🌪️ The Volatility Trader Only trades when volatility explodes — bigger swings, bigger profits	+19.2%	62.5%	72(~1/mo)	0%	—
16	🇺🇸 The Trump Trader Trades when the president tweets	+8.5%	100%	1(~1/1yr)	0%	100%
17	🌊 The Calm Before Storm Buys when volatility is at its lowest — the calm before the explosion	+0.1%	44.4%	18(~1/4mo)	0%	78%
18	🐝 The Hive Mind Follows Reddit sentiment	-1.3%	50%	4(~1/4mo)	-1.3%	38%
19	↔️ The Divergence Bot When price and sentiment disagree, something is about to happen	-8.5%	50%	4(~1/1yr)	-8.5%	60%
20	⛏️ The Hash Rate Bot Follow the miners — they know something you don't	-9.8%	36.8%	19(~1/2mo)	-9.8%	—
21	📊 The VWAP Bot Buys below the 20-day average, sells above	-12.5%	71.9%	57(~1/2mo)	-12.5%	—
22	🥧 The Pi Bot Trading with mathematical constants. The ultimate placebo.	-26.9%	51%	49(~1/2mo)	-26.9%	17%
23	🔊 The Volume Spike Buys when volume explodes and price dips — Wyckoff accumulation	-27.6%	48.6%	37(~1/2mo)	-27.6%	100%

All experiments

Bot details

Sorted by return. Click any card for full strategy details.

⏳

The Halving Bot

Buy 6 months before halving, sell 12 months after. +91%.

Running

BTCJan 2020~1 every 38 months

⏳ Bitcoin halving schedule + daily pricesBacktest since Jan 2020 · Live since Apr 2026

$10k →$220,415.62

Jan 20Today

+2104.2%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

100%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 38 months)?

2 completed round-trip trades since Jan 2020. Frequency: ~1 every 38 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

100%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

UNDENIABLE — +91% on a single trade. The halving cycle is the most predictable pattern in crypto. But with 1 trade per 4 years, it is a meta-signal, not a trading strategy. The real value: it tells you WHEN to accumulate. Next actionable date: October 2027. Use this as a timing overlay for other strategies, not standalone.

📆

The Seasonality Bot

Sell in May? We tested crypto seasonality.

Running

BTCJan 2020~1 every 11 months

📆 Calendar-based (no external data)Backtest since Jan 2020 · Live since Apr 2026

$10k →$158,276.59

Jan 20Today

+1482.8%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

71.4%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 11 months)?

7 completed round-trip trades since Jan 2020. Frequency: ~1 every 11 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

100%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

INTERESTING — Avoided the 2024 summer chop, caught the post-election rally (+58.6%). But only 2.3 yearly cycles in the data, too few for statistical significance. The concept of crypto seasonality is real but fragile. Would need 5+ years of data to confirm. Still: being in cash May-Sep saved significant pain.

🔄

The Contrarian

Does the OPPOSITE of what RSI says. And it works.

Running

BTCJan 2020~1 every 8 months

🔄 RSI(14) inverted signals (calculated locally)Backtest since Jan 2020 · Live since Apr 2026

$10k →$55,387.48

Jan 20Today

+453.9%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

55.6%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 8 months)?

9 completed round-trip trades since Jan 2020. Frequency: ~1 every 8 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

100%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

PROVOCATIVE — Proves that the standard RSI playbook is WRONG for BTC. Buying at RSI>70 catches strong trends early. Selling at RSI<30 exits before capitulation. Two massive winners (+27.7%, +28.3%) and one small loser (-5%). Compare to our standard RSI Bot (+0.2%) doing the textbook version. The lesson: in trending markets, strength begets strength.

😱

The Panic Buyer

Buys when everyone screams SELL

Running

BTCJan 2020~1 every 19 months

📊 Fear & Greed API (alternative.me)Backtest since Jan 2020 · Live since Apr 2026

$10k →$49,100

Jan 20Today

+391%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

50%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 19 months)?

4 completed round-trip trades since Jan 2020. Frequency: ~1 every 19 months.

-28%

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

60%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

🔴 FRAGILE — Extreme Fear (F&G below 15) happens very rarely. Only 1 completed trade in 16 months. The single entry at F&G=10 (Feb 2025) was badly timed — it caught a falling knife. The thesis is correct (buy fear, sell greed) but the execution is too blunt. A multi-signal approach combining F&G with trend confirmation would likely work better.

🌕

The Lunatic

Full moon buy, new moon sell. Yes, really.

Running

BTCJan 2020~1/month

🌕 NASA lunar data (PyEphem)Backtest since Jan 2020 · Live since Apr 2026

$10k →$31,479.07

Jan 20Today

+214.8%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

51.3%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1/month)?

78 completed round-trip trades since Jan 2020. Frequency: ~1/month.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

🟡 BASELINE — The control experiment did its job. At -9.1%, the moon performs roughly like random trading minus transaction costs. Any real strategy must beat this to prove it has an edge. Interestingly, 50% win rate is exactly what you would expect from pure chance.

🌀

The Fibonacci Bot

Sacred geometry or trader astrology? We tested it.

Running

BTCMar 2020~1 every 5 months

🌀 Fibonacci retracement from daily swing pointsBacktest since Mar 2020 · Live since Apr 2026

$10k →$25,551.28

Mar 20Today

+155.5%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

80%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 5 months)?

15 completed round-trip trades since Mar 2020. Frequency: ~1 every 5 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

DEBUNKED — Fibonacci retracement is officially trader astrology. It works in bull markets (the bounces from 61.8% are real when trend is up) but fails catastrophically in bears (price blows through the level like it does not exist). The 60% win rate looks decent, but the losers are bigger than the winners. Verdict: coincidence detector, not predictive tool.

🏃

The Momentum Bot

Rides winning streaks, exits on first red day

Running

BTCJan 2020~1/month

📊 BTC daily prices (CryptoCompare, free)Backtest since Jan 2020 · Live since Apr 2026

$10k →$22,996.69

Jan 20Today

+130%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

35.9%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

128

Trades (~1/month)?

128 completed round-trip trades since Jan 2020. Frequency: ~1/month.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

83%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

STRONG — Despite only 37% win rate, the winning trades are much larger than the losers. Momentum works because crypto trends are real: when BTC starts moving, it often continues for days. The 13.9% fee drag is significant but the strategy overcomes it. A serious graduation candidate.

🧬

The Entropy Bot

Uses information theory to detect regime changes. Beats HODL.

Running

BTCMay 2020~1 every 3 months

🧬 Shannon entropy from daily returns (calculated locally)Backtest since May 2020 · Live since Apr 2026

$10k →$21,754.15

May 20Today

+117.5%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

60.9%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 3 months)?

23 completed round-trip trades since May 2020. Frequency: ~1 every 3 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

100%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

BREAKTHROUGH — The only bot that BEATS Buy & Hold (+83.1% vs +75.4%). Shannon entropy detects something no other indicator does: when returns become STRUCTURED (low entropy = trend forming) vs RANDOM (high entropy = noise). 85.7% win rate with only 7 trades. Information theory works as a trading signal. A top graduation candidate.

🐢

The Turtle

Classic Donchian channel breakout — the original turtle trading

Running

BTCJan 2020~1/month

📊 BTC daily OHLCV (CryptoCompare, free)Backtest since Jan 2020 · Live since Apr 2026

$10k →$20,224.37

Jan 20Today

+102.2%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

47.4%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1/month)?

95 completed round-trip trades since Jan 2020. Frequency: ~1/month.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

100%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

SOLID — Classic trend-following works on crypto. The Donchian breakout catches every major trend but gets whipsawed in choppy sideways markets. 33 trades with 8.3% fee drag is manageable. The ATR trailing stop helps protect profits but sometimes exits too early. A proven methodology with 40+ years of track record.

📉

The RSI Bot

The most popular indicator. Does it actually work?

DoneBusted

BTCMar 2020~1 every 8 months

📉 RSI calculated from daily prices (free)Backtest since Mar 2020 · Live since Apr 2026

$10k →$14,645.32

Mar 20Today

+46.5%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

66.7%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 8 months)?

9 completed round-trip trades since Mar 2020. Frequency: ~1 every 8 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

SURPRISING — RSI performed much better over 6 years (+46.5%) than over 16 months (+0.2%). The 2020-2021 bull market produced strong RSI signals. Still underperforms HODL (+931%), but not the failure we expected. RSI works better in strong trends than in choppy markets.

📅

The Weekender

Monday buy, Friday sell. Every week.

Running

BTCJan 2020~1/week

⏰ Calendar-based (no external data)Backtest since Jan 2020 · Live since Apr 2026

$10k →$13,005.96

Jan 20Today

+30.1%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

51.5%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

328

Trades (~1/week)?

328 completed round-trip trades since Jan 2020. Frequency: ~1/week.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

MODERATE — Over 6 years, the Weekender made +30.1%. Not because of a weekday effect, but because being in BTC 5 out of 7 days during a 10x appreciation still captures most of the move. The strategy has no real edge, it just benefits from overall market growth.

⚡

The Scalper

Buys every dip, sells every bounce. Speed kills.

Running

BTCJan 2020~1 every 2 weeks

📊 BTC daily prices (CryptoCompare, free)Backtest since Jan 2020 · Live since Apr 2026

$10k →$12,999.79

Jan 20Today

+30%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

68.9%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

177

Trades (~1 every 2 weeks)?

177 completed round-trip trades since Jan 2020. Frequency: ~1 every 2 weeks.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

REDEEMED — Over 6 years, the Scalper actually made +30.0% despite heavy fee drag. More dip-recovery cycles in the longer dataset gave the strategy enough edge to overcome fees. Still, the return-to-risk ratio is poor compared to simpler strategies.

🐋

The Whale Watcher

Follows the smart money on-chain

Running

BTCJan 2024~1 every 6 months

⛓️ Blockchain.com API + Mempool.space (both free)Backtest since Jan 2024 · Live since Apr 2026

$10k →$12,878.59

Jan 24Today

+28.8%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

60%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 6 months)?

5 completed round-trip trades since Jan 2024. Frequency: ~1 every 6 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

100%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

🟢 WORKS — On-chain transaction volume is a real signal. When whales move large amounts of BTC, prices follow within days. The 90-day percentile approach avoids false signals from normal volume fluctuations. Key insight: volume spikes PRECEDE price moves, not the other way around. This bot is a serious graduation candidate.

📈

The Funding Rate Bot

Trades against the crowd when leverage gets extreme

Running

BTCJan 2024~1 every 14 months

📈 Bybit API (free, real-time funding rates)Backtest since Jan 2024 · Live since Apr 2026

$10k →$12,788.5

Jan 24Today

+27.9%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

50%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 14 months)?

2 completed round-trip trades since Jan 2024. Frequency: ~1 every 14 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

🟢 PROMISING — Extreme funding rates signal crowded positioning. When everyone is leveraged long, a squeeze is coming. The first trade (+43%) caught the July-November 2024 rally perfectly. Only 2 trades in 16 months means the bot is very selective. Needs more data to confirm, but the logic is sound and proven in our live S37 strategy.

🌪️

The Volatility Trader

Only trades when volatility explodes — bigger swings, bigger profits

Running

BTCMar 2020~1/month

🌪️ BTC daily OHLCV + ATR volatility (free)Backtest since Mar 2020 · Live since Apr 2026

$10k →$11,915.56

Mar 20Today

+19.2%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

62.5%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1/month)?

72 completed round-trip trades since Mar 2020. Frequency: ~1/month.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

MIXED — High win rate (64%) and the volatility filter successfully avoids low-vol chop. But got crushed in the Nov-Dec 2025 persistent crash: four consecutive stop-losses. The bot works in V-shaped dips but fails in slow grinding bear markets. Needs an additional trend filter to avoid catching falling knives in downtrends.

🇺🇸

The Trump Trader

Trades when the president tweets

Running

BTCJan 2025~1 every 15 months

📡 GDELT News API (free, 3B+ articles)Backtest since Jan 2025 · Live since Apr 2026

$10k →$10,846.24

Jan 25Today

+8.5%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

100%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 15 months)?

1 completed round-trip trades since Jan 2025. Frequency: ~1 every 15 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

100%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

DISAPPOINTING — Keyword-based sentiment scoring is too crude for nuanced political statements. The GDELT data captures volume of coverage but not quality. A headline like Trump crypto regulation fears scores as bearish even if the article explains why regulation is actually bullish. GPT-4o classification would likely improve results significantly.

🌊

The Calm Before Storm

Buys when volatility is at its lowest — the calm before the explosion

Running

BTCJun 2020~1 every 4 months

🌊 ATR volatility percentile (calculated locally)Backtest since Jun 2020 · Live since Apr 2026

$10k →$10,014.21

Jun 20Today

+0.1%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

44.4%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 4 months)?

18 completed round-trip trades since Jun 2020. Frequency: ~1 every 4 months.

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

78%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

SOUND THESIS — Vol compression is a real phenomenon. The Oct 2024 entry (+26.4%) and Apr 2025 entry (+15.0%) prove the concept works in the right regime. Near break-even despite a bear phase. The problem is not timing, it is DIRECTION. Needs a stronger trend filter. Worth iterating on.

🐝

The Hive Mind

Follows Reddit sentiment

Running

BTCFeb 2025~1 every 4 months

🐝 GDELT social data + Reddit JSON API (both free)Backtest since Feb 2025 · Live since Apr 2026

$10k →$9,869.89

Feb 25Today

-1.3%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

50%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 4 months)?

4 completed round-trip trades since Feb 2025. Frequency: ~1 every 4 months.

-1.3%

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

38%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

WEAK SIGNAL — Near zero return (-1.3%) over the full period. Crowd sentiment remains a lagging indicator. GDELT social data as Reddit proxy is too noisy for reliable signals. Confirms our earlier research on sentiment indicators.

↔️

The Divergence Bot

When price and sentiment disagree, something is about to happen

Running

BTCJan 2024~1 every 7 months

↔️ F&G API + price divergence (free)Backtest since Jan 2024 · Live since Apr 2026

$10k →$9,150

Jan 24Today

Live data collecting — check back soon

-8.5%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

50%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 7 months)?

4 completed round-trip trades since Jan 2024. Frequency: ~1 every 7 months.

-8.5%

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

60%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

CONFIRMS WEAKNESS — Fear & Greed divergence signals are too rare (7 days out of 835) and inconsistent. This confirms our earlier research: F&G measures sentiment AFTER the fact, not before. The signal is theoretically appealing but practically useless for timing.

⛏️

The Hash Rate Bot

Follow the miners — they know something you don't

Running

BTCJan 2024~1 every 2 months

⛏️ Blockchain.com hash rate API (free)Backtest since Jan 2024 · Live since Apr 2026

$10k →$9,020

Jan 24Today

Live data collecting — check back soon

-9.8%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

36.8%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 2 months)?

19 completed round-trip trades since Jan 2024. Frequency: ~1 every 2 months.

-9.8%

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

NOISY SIGNAL — Hash rate has structural lag: by the time miners deploy hardware, the price move is often already priced in. Some excellent individual trades (+17%, +11.5%, +10.2%) but too many false signals. The signal is fundamentally sound but needs more sophisticated filtering like hash ribbons or MA crossovers.

📊

The VWAP Bot

Buys below the 20-day average, sells above

Running

BTCFeb 2020~1 every 2 months

📊 BTC daily prices (CryptoCompare, free)Backtest since Feb 2020 · Live since Apr 2026

$10k →$8,746.87

Feb 20Today

-12.5%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

71.9%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 2 months)?

57 completed round-trip trades since Feb 2020. Frequency: ~1 every 2 months.

-12.5%

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

OVERFITTING — Was +15.0% on 16 months, dropped to -12.5% on 6 years. Mean reversion around a 20-day MA fails in strong trends: buys too early in crashes, sells too early in rallies. The 2022 bear market destroyed this strategy.

🥧

The Pi Bot

Trading with mathematical constants. The ultimate placebo.

Running

BTCJan 2020~1 every 2 months

🥧 Mathematical constants in price digits (local calculation)Backtest since Jan 2020 · Live since Apr 2026

$10k →$7,310.92

Jan 20Today

-26.9%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

51%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 2 months)?

49 completed round-trip trades since Jan 2020. Frequency: ~1 every 2 months.

-26.9%

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

17%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

PERFECT PLACEBO — Lost -22.8%, exactly what random noise + fees should produce. This is the BotLab's most important bot: it proves that without real edge, you just pay fees. If your strategy cannot beat the Pi Bot, it has no edge. Pattern appeared on only 1.8% of trading days.

🔊

The Volume Spike

Buys when volume explodes and price dips — Wyckoff accumulation

Running

BTCMar 2020~1 every 2 months

🔊 BTC daily volume + price (CryptoCompare, free)Backtest since Mar 2020 · Live since Apr 2026

$10k →$7,239.4

Mar 20Today

-27.6%

Return?

Total profit or loss since the bot started, as a percentage of the initial $10,000.

48.6%

Win Rate?

Percentage of trades that were profitable. 50% means half the trades made money.

Trades (~1 every 2 months)?

37 completed round-trip trades since Mar 2020. Frequency: ~1 every 2 months.

-27.6%

Max DD?

Maximum Drawdown: the worst peak-to-trough decline. If this is -15%, the portfolio dropped 15% from its highest point before recovering.

100%

Robust?

Parameter robustness: how many parameter variants are profitable. 100% = works with any setting (robust). Below 40% = only works with one specific tuning (fragile).

OVERFITTING EXPOSED — Was +45.7% on 16 months, collapsed to -27.6% on 6 years. The volume spike pattern that worked in 2024-2025 failed in the 2022 bear market: high-volume dips were followed by MORE dips, not recoveries. The poster child for why short backtests are dangerous. Still running as a cautionary example.

Your idea, our lab

Got a strategy idea?

Tell us what to test next. Subscribe to the newsletter and reply with your craziest strategy idea. If it's testable, we'll build it.

Free

Bot Alerts & Trading Lies

Get notified instantly when the bot buys or sells. Plus: free PDF, weekly myth-busting and bot performance updates.

Bot Signal AlertsFree PDF

No spamUnsubscribe anytimeYour data stays with us

Graduated from the lab

5 bots that cleared the bar

The Watchdog on real capital. Tactician, Genius and Alpha Hunter on paper. Scout on backtest-only. Tier-labeled, walk-forward tested, honest about what's live.

Meet the live bots

Experiments that got retired

And the ones that didn't clear the bar

Every failed walk-forward, every downgraded bot, every retired experiment. Documented publicly. So you can trust the survivors.

See the failures

Disclaimer: The Bot Lab uses virtual money only. No real trades are executed. Past performance (including backtests) does not guarantee future results. This is an educational experiment, not financial advice.