YouTube trading educator (video S2HaCa0b-bY, 'MACD Money Map'), 2025
Trend Following
Evidence: debunked_botty
swing
cryptoforex
2/10
Relevance for Botty
YouTube strategy (S2HaCa0b-bY): MACD decomposed into 3 'systems' — trend (zero-line bias + crossover + distance rule + delay), reversal (divergence + histogram), confirmation (triple timeframe + price action). Tested at Botty and killed as a single-asset TA curve fit.
Core concept
Standard MACD (12/26/9) as three combined systems instead of 'random crossovers': (1) Trend — daily MACD sign as an absolute bias (>0 longs only, <0 shorts only), 4h crossover only 'far from 0' (distance rule), 2-3 candle delay; (2) Reversal — classic MACD divergence + histogram patterns (flip/shrinking/zero-bounce); (3) Confirmation — triple-timeframe stack (daily bias / 4h setup / 1h trigger, 4x multiplier) + price action (support/hammer/trendline). Execution: entry@close, swing stop, 2R target, half off + BE stop + trail on the opposite crossover.
Relevance for Botty
Relevance Score 2/10
TESTED 2026-06-24 (backtesting/macd_money_map_scout.py) — faithful mechanizable core: daily zero-line bias + 4h crossover + distance rule (scale-free k*std(MACD)) + N-candle delay + swing stop + 2R/half-off/trail. BTC 4h 2020-2026, after costs (3.5bps/leg). (1) BTC SWEEP over k={0,0.5,1.0} × delay={0,2,3}: ALL 9 configs negative (CAGR -6.8% to -21.5%), real win rate ~30% (not 60%), maxDD -62% to -86%, per-trade Sharpe negative everywhere. The only positive year is almost always 2020 (COVID bull). The '2-3 candle delay' makes several configs worse. (2) CROSS-ASSET GATE (frozen k=0.5/D=2): BTC -18.7%, ETH -13.9%, SOL +2.9% (only via variance, +155bps avg at -84% maxDD) → NOT consistent = BTC/asset curve fit, no structural edge. Consistent with divergences_dead + donchian@2h kill + TSMOM kill. NOT tested mechanically (deliberately): divergence (already dead 3x) + price-action confirm (discretionary — that is exactly where curve-fit narratives hide). RESCUE TEST of the missing (filter) parts (macd_money_map_filters.py): most of the missing pieces are filters = they only select subsets of the trades. Tested: support proxy (tight-risk), hammer wick, hard distance, real 1h histogram trigger (BTC only). Result: NO filter is cross-asset consistent — tight-risk rescues ETH (+111bps) but not BTC, the 1h trigger rescues BTC (+103bps, n=41, ~7 trades/yr) but is not verifiable on ETH/SOL, hammer is negative everywhere (BTC -135bps). That DIFFERENT filters rescue DIFFERENT assets = the signature of multiple testing/overfit (~18 subsets tried), not edge. CONCLUSION: typical YouTube TA, fails at the cheap gates; even the missing parts don't rescue it.
Rules
Entry
- Daily MACD > 0 → longs only / < 0 → shorts only (zero line as an 'absolute law')
- 4h bullish/bearish MACD crossover in the direction of the bias
- Distance rule: the crossover must be 'far from 0' (video: |MACD|>0.5 — forex pip scale)
- Wait 2-3 candles after the crossover, then enter at the candle close
- Optional triple-timeframe confirmation (1h histogram flip as the trigger)
Exit
- Stop at the most recent swing high/low
- Target 2R: take half off there, move stop to breakeven
- The rest trails until the opposite MACD crossover
Parameter
| Name | Typ. value | Description |
|---|---|---|
| macd | 12/26/9 | Standard MACD |
| distance_rule | |MACD|>0.5 (Forex) | Price-scale dependent — translated scale-free to k*std(MACD) on BTC |
| confirm_delay | 2-3 candles | Wait time after the crossover |
| target | 2R | Risk-reward target |
Pros & Cons
Pros
- Clearly mechanizable trend core (system 1 + 3-stack)
- Zero-line bias + distance rule are sensible trend-filter ideas
- Low frequency, unambiguous rules
Cons
- Tested at Botty: negative on BTC across ALL parameters (see botty_applicability)
- The distance rule '0.5' is forex-pip scaled, meaningless on crypto without normalization
- The reversal system (divergence) was debunked 3x walk-forward at Botty
- Price-action confirmation is discretionary — not reproducibly testable
- Marketing claims (60% win rate, 'prints money') don't hold out-of-sample with costs
Bot suitability
Poor — tested and killed. Single-asset BTC TA family that Botty's gates repeatedly and correctly expose as overfit.