RESEARCH LAB — READ ONLY

Strategy Research

Before going back to real capital, every strategy gets tested here first. These are the actual AI reasoning traces and backtest results — published openly so you can follow the logic, spot flaws, and learn from what works and what doesn't.

Research runs are triggered privately. Use the comment box to ask questions or flag issues — feedback is read and shapes future research, but does not trigger AI automatically.

NEEDS MORE DATAbreakoutmomentumBTCtier3entry-signal

Does tbot's Tier 3 breakout signal actually have edge? (BTC, 6 months)

Hypothesis: tbot's current Tier 3 entry: 3 consecutive up-days + volume spike >1.2x 7-day average. Testing whether this condition has statistically significant positive expectancy on BTC daily data.

11 Aug 2026

Win Rate60.0%

Sharpe-1.49

Trades5

Timeframe6 months

AssetsBTC/USDT

Research output

## Strategy
Entry: BUY when (1) 3 consecutive days of positive close-to-close returns AND (2) today's volume > 1.2x the 7-day average volume.
Exit: Fixed 5-day hold, exit at close.

## Trade Log

| # | Entry Date | Exit Date | Entry Price | Exit Price | Return % | Hold | W/L |
|---|---|---|---|---|---|---|---|
| 1 | 2026-03-14 | 2026-03-19 | $70,965 | $71,211 | +0.35% | 5d | ✓ |
| 2 | 2026-04-18 | 2026-04-28 | $77,111 | $77,333 | +0.29% | 10d* | ✓ |
| 3 | 2026-05-07 | 2026-05-12 | $81,488 | $81,705 | +0.27% | 5d | ✓ |
| 4 | 2026-07-07 | 2026-07-12 | $64,075 | $63,787 | -0.45% | 5d | ✗ |
| 5 | 2026-07-31 | 2026-08-05 | $64,812 | $64,074 | -1.14% | 5d | ✗ |

*Trade 2 held 10 days — a second signal fired while already in position.

## Results Summary

| Metric | Value |
|---|---|
| Total trades | 5 |
| Win rate | 60% |
| Avg return per trade | -0.14% |
| Profit factor | 0.57 (losses outweigh wins) |
| Annualised Sharpe | -1.49 |
| vs Buy-and-Hold | +3.19% (avoided the July drawdown) |

## Monte Carlo Permutation Test (1,000 simulations)

| Metric | Actual | Simulated Mean | p-value |
|---|---|---|---|
| Sharpe | -6.96 | -2.72 | 0.978 |
| Max Drawdown | -1.57% | — | 0.805 |

## Verdict: NEEDS MORE DATA

With only 5 trades in 6 months, the p-values (0.978 and 0.805) are far above the 0.05 significance threshold. The results are statistically indistinguishable from random trade ordering. A minimum of 30-50 trades (roughly 3-5 years of data) would be needed to draw reliable conclusions.

What we can say:
- 60% win rate sounds good but losses were larger than wins (profit factor 0.57)
- The strategy avoided being in BTC during the July-August drawdown — but with only 2 losing trades that's not a pattern, it's luck
- The entry condition is too rare on daily BTC data — fires ~10x/year

## Suggested Next Steps
1. Relax to 2 consecutive up-days (more signals, same volume filter)
2. Add a 50-day MA trend filter — only trade when price is above MA50
3. Extend backtest to 2-3 years to get a statistically meaningful sample
4. Test on altcoin pairs where the breakout pattern may be more frequent

Questions & Feedback (0)