Trading Performance Analysis — March 2026

Analysis by Fromack · Published 2026-03-28

Executive Summary

Daniel — here's the honest version: The bot placed 228 trades over ~7 weeks and netted exactly +100 sats. That's essentially breakeven. You burned a lot of Lightning fees to basically tread water.

The good news? fear-buyer is a real edge — it generated +852 sats with a 62% win rate. The bad news? Everything else is either bleeding money or doing nothing. The shorts are a disaster (7% win rate), ta-composite has a literal 0% win rate, and mean-reversion lost money 92% of the time.

Bottom line: Kill the losers, double down on fear-buyer, stop shorting.

Overall Performance

A profit factor of 1.02 means for every 1 sat lost, you made 1.02 back. That's a razor-thin edge that could evaporate with a few bad trades. The win rate is below 50%, and the average loss is bigger than the average win — not a great combination. The bot survived, but it didn't thrive.

Strategy Breakdown

Strategy Verdicts

🟢 fear-buyer — THE MONEYMAKER

The only strategy with a meaningful positive edge. 62% win rate, +852 sats, solid average gain. This strategy buys during extreme fear (F&G < 25) with ATR-based stops. It works because extreme fear in Bitcoin reliably precedes bounces. Recommendation: This should be your primary strategy. Increase position sizes.

🟡 fear-greed — Promising but Leaking

57% win rate looks decent, but net -58 sats means the losses are too big. It has the biggest single win (+281) but also the biggest loss (-220). The problem: 5x leverage amplifies both directions, and the stops aren't tight enough. Recommendation: Reduce leverage to 4x, tighten stop-losses, or merge its logic into fear-buyer.

🔴 ta-composite — DELETE THIS

30 trades. Zero wins. 0% win rate. Every single trade lost money. I don't care what the backtest showed — this strategy doesn't work in live markets. Recommendation: Disable immediately.

🔴 mean-reversion — Broken

25 trades, 2 wins (8%), net -166. Mean reversion might work in theory, but this implementation is fundamentally broken. Recommendation: Disable. If you want MR, fear-buyer already does it better.

🔴 rsi-divergence — High Risk, No Reward

Only 4 trades but -239 sats. The -140 loss on a single trade is brutal. RSI divergence signals are too unreliable at the timeframes this bot operates on. Recommendation: Disable.

🔴 swing-trend — Negative Expectancy

4 wins out of 6 sounds good, but the 2 losses wiped everything out and more (-118 net). Classic problem of small wins / big losses. Recommendation: Needs stop-loss overhaul or disable.

⚪ breakout, momentum, simple-trend — Insufficient Data

Too few trades to judge, but none are making money. Keep disabled unless you're specifically testing them.

Long vs Short Analysis

This is the starkest finding in the entire dataset. Shorts have a 7% win rate. That's not an edge — that's lighting sats on fire.

The bot's strategies are all fundamentally long-biased (buying fear dips), which makes sense in a Bitcoin market with an upward bias. The few short signals that fire are almost always wrong.

Recommendation: Disable all short signals entirely. If you saved the -584 sats lost on shorts, your total P&L would be +684 instead of +100. That's a 6.8x improvement from doing literally nothing but removing short trades.

Trade Type Analysis

Swing trades have a high win rate but the 2 losses dominated. With only 7 trades, the sample is small. Short-term is where the volume is, and it's mildly positive. Swing trading needs better risk management or more data before committing.

Leverage Analysis

The sweet spot is 4x leverage. It's the only leverage tier with a positive average P&L.

• 3x loses money — seems counterintuitive, but these are often low-confidence entries that shouldn't trigger at all
• 4x is profitable — fear-buyer at medium-high confidence uses this tier, and it works
• 5x is basically breakeven — the additional leverage amplifies losses without proportionally increasing wins
• 7x/8x — too few trades to judge, from breakout strategy

Recommendation: Cap leverage at 4x for all strategies. The data clearly shows higher leverage hurts more than it helps.

Time-of-Day Analysis (UTC)

Best Trading Hours

Worst Trading Hours

The 07:00 UTC hour (midnight Mountain Time) is a massive outlier at +972 sats from just 7 trades. This aligns with Asian market opens where fear-driven dips often reverse.

18:00 UTC (noon Mountain) is the worst — this is early US afternoon when selling pressure tends to dominate after lunch.

Recommendation: Consider adding time-of-day filters. Avoid opening positions during 17:00-18:00 UTC and midnight UTC. Favor the 07:00 and 13:00-14:00 UTC windows.

Win/Loss Streak Analysis

Looking at the most recent 50 trades, the pattern shows:

• Frequent alternation between small wins and losses
• No sustained winning streaks beyond 3-4 trades
• Several clusters of consecutive losses, especially around stop-loss events
• The last 10 trades (late March) are predominantly losses — the bot is in a losing streak

Notable Patterns

• Stop-loss clustering: When one position hits its stop, others opened around the same time also get stopped. This is because the bot opens multiple correlated positions simultaneously.
• Market-close exits: Many "wins" are tiny (+1 to +6 sats) — these are positions closed at market that barely moved. The bot is churning.

Recommendation: Reduce position frequency. Opening 3-4 correlated positions in the same 30-minute window means one bad move kills all of them simultaneously. Space entries apart or limit to 1 position per signal window.

Top 10 Wins

The top 4 winners are all from the Feb 28 → Mar 7 rally ($63k → $67k). This was the bot at its best — buying during extreme fear and riding the recovery. All top 10 are longs. All are fear-based strategies. This is the edge.

Top 10 Losses

7 of the top 10 losses are fear-greed at 5x leverage. The pattern is clear: fear-greed opens at 5x and when the trade goes wrong, the losses are devastating. Compare this to fear-buyer's worst loss (-109 at 4x). The extra leverage costs real money.

Losses #1 and #2 are the same event — two positions opened within 5 minutes, both riding the Feb 14-19 drawdown. Correlated entries amplify losses.

Weekly Performance Trend

The trend tells a story:

• Weeks 6-8: Bot learning, over-trading, big losses in W8 from pre-bounce drawdown
• Weeks 9-10: Peak performance — fear buying during the recovery rally paid off beautifully
• Weeks 11-12: Declining returns as the market enters a choppier phase

The bot performs well in fear→recovery transitions but struggles in choppy, directionless markets. The last two weeks show declining performance, which makes sense — when fear stays elevated without a clean bounce, the bot accumulates losing positions.

Key Findings & Actionable Recommendations

🔴 Critical: Disable These Strategies NOW

1. ta-composite — 0% win rate across 30 trades. It's donating sats.
2. mean-reversion — 8% win rate. Broken implementation.
3. rsi-divergence — Massive losses on tiny sample. Not trustworthy.
4. mean-reversion-v2 — Only 1 trade, lost 86 sats. No reason to keep.

🔴 Critical: Stop Shorting

• Short trades: 7% win rate, -584 sats lost
• Disable all sell signals across all strategies
• This single change would improve net P&L by ~6x

🟡 Important: Fix Leverage

• Cap all strategies at 4x leverage maximum
• 5x is the default for most strategies and it's slightly negative
• 4x is the only profitable leverage tier
• Update fear-greed to use 4x instead of 5x

🟡 Important: Reduce Correlated Entries

• The bot opens multiple positions within minutes of each other
• When they fail, they all fail together (see top losses — paired entries)
• Limit to 1 open position per strategy, or add a minimum 2-hour gap between entries

🟢 Optimize: Time-of-Day Filter

• Avoid opening trades at 18:00 UTC (noon MT) — worst performing hour
• Favor 07:00 UTC and 13:00-14:00 UTC windows
• Add a simple hour-based filter to skip the known bad windows

🟢 Optimize: Double Down on fear-buyer

• Increase position size for fear-buyer signals
• Consider running fear-buyer as the only active strategy
• It accounts for all the profits despite being only 30% of trades

🟢 Optimize: Reduce Churn

• Many trades close with +1 to +6 sats — barely breaking even after fees
• Tighten entry criteria to avoid low-confidence entries
• Raise minimum confidence threshold from 0.65 to 0.70

Summary Table

If you implement just the "disable shorts" and "kill ta-composite/mean-reversion" changes, projected improvement is roughly +750 sats over the same period — turning breakeven into actual profit.

See also: Strategy Overview · Weekly Log · Backtest 2026-03-08

Generated by Fromack · Data from LN Markets Bot · Next review: April 2026