Experiments
Every question we've asked of the data, with the answer, including the dull ones and the failures. 79 studies so far: 7 supported, 35 not supported, 22 inconclusive, and 15 where the work was done but the results weren't saved (we say so rather than guess).
How to read these. "Supported" means the evidence backs the idea; "not supported" means it doesn't, which is a result, not a failure. Each study also has a trust rating: TradingView's own fills on large samples rate high; our early Python simulations rate low, because several of them knew in hindsight where the pattern's turning point was, which made results look up to three times better than reality.
Entries · Exits · Filters · Patterns · Timeframes · Instruments & watchlist · Market regime · Sizing · Machine learning · Outside indicators & papers
Entries (9)
- Enter at D+1 (idealised) or at real-time confirmation? No saved results · trust: low
- Would a limit order at Point D (or a 50% pullback) beat the confirmation entry? No saved results · trust: low
- Four 'improvements' on a 113-stock 1h basket No saved results · trust: low
- Fixing our random-entry control Supported · trust: medium
- Does the random control understate the harmonic edge because it overlaps the pattern itself? Not supported · trust: medium
- What should 'random entry' mean? A price-blind reference control Supported · trust: medium
- Does Harmonic Master's entry beat a random entry with the same stop and target? (updated 2026-10-11, price-blind reference) Not supported · trust: medium
- Does a harmonic pattern add anything beyond buying a 15% pullback? Inconclusive · trust: medium
- Where the harmonic backtest sits against random buying and SPY Not supported · trust: medium
Exits (11)
- Are the 1.5R / 2R targets too far? Does breakeven help? Inconclusive · trust: medium
- Move the stop to breakeven at Point A or C instead of at TP1? Not supported · trust: low
- Should the runner aim for 3R when D touches Camarilla S3? No saved results · trust: low
- Exit on the first red Renko brick instead of fixed targets? Inconclusive · trust: low
- Is a tight stop just under D (D x 0.998) better than min(D, X) x 0.99? Not supported · trust: medium
- Do setups with tighter stops (in % of price) do better? Not supported · trust: high
- Swing-low stop with the prior 24-month high as target? Inconclusive · trust: medium
- Would bigger fixed targets (3R, 5R) beat 2R? Not supported · trust: medium
- Enter on a SuperTrend flip after the pattern and trail the stop with SuperTrend? No saved results · trust: low
- Is the dual-tranche exit (50% at 1.5R, runner to 2R) better than a single target? Inconclusive · trust: low
- Does moving to breakeven after TP1 help on 1h, 4h and daily? Supported · trust: high
Filters (13)
- Does any indicator filter (RSI, MAs, Kalman VWAP, SuperTrend, daily pivots) improve 15m setups? Not supported · trust: high
- Does an 'RSI <= 40 at Point D' filter help on 1h? Not supported · trust: high
- Do RSI, MA or RSI-divergence conditions help on any timeframe? (catalogue re-check) Inconclusive · trust: medium
- Do setups whose D sits on a Renko support shelf do better? Inconclusive · trust: low
- Does Camarilla S3 confluence at D improve setups? No saved results · trust: low
- Does volume-profile (FRVP/POC) confluence improve setups? No saved results · trust: low
- Only buy when D sits at the 50-61.8% retracement of the big swing? Not supported · trust: medium
- Does the Fib-zone idea work on mega caps only? Not supported · trust: low
- Does the Bulls v Bears (BvB v6) filter improve the daily top-40? Not supported · trust: low
- Inside falling markets, does any extra filter help? Inconclusive · trust: medium
- Do more volatile instruments do better on 15m? Inconclusive · trust: low
- Do more volatile stocks or setups do better, on any timeframe? Not supported · trust: medium
- Do harmonics fail when a stock's recent daily returns have been trending (positive autocorrelation)? Inconclusive · trust: low
Patterns (8)
- Are some harmonic patterns better than others on 15m? Not supported · trust: high
- Does the pattern ranking hold across timeframes? Not supported · trust: medium
- Are Bat (100%) and Deep Crab (91.7%) 'S-tier' patterns? Not supported · trust: low
- Does removing the Standard Crab improve the hand-picked list? Not supported · trust: low
- Is 'Gartley in an uptrend' a real edge? Not supported · trust: medium
- Would a smaller zigzag give more setups without losing edge? Not supported · trust: medium
- Zigzag length 12 vs 25 vs 50 (LuxAlgo-style) No saved results · trust: low
- Do patterns with AB = CD symmetry (equal legs in price and time) do better? Not supported · trust: medium
Timeframes (6)
- Does the unfiltered strategy make money on every timeframe, and which earns most per setup? Supported · trust: high
- Does the strategy work on weekly charts? Inconclusive · trust: low
- Is 1h '4.7x better' than daily? Not supported · trust: low
- Do the best 1h instruments also work on 15m and 4h? Not supported · trust: low
- First Python backtest: how good are harmonic setups on 1h/4h/1d/1W? Not supported · trust: low
- Weekly vs daily, and small vs medium vs large patterns No saved results · trust: low
Instruments & watchlist (14)
- Are some stocks reliably better harmonic stocks than others? Not supported · trust: high
- Does picking the best-backtesting stocks give a better live basket? Not supported · trust: high
- Early full-universe sweeps (1h? and daily, full history) Inconclusive · trust: low
- Does the daily top-40 list hold up over 2016-2026? Inconclusive · trust: low
- min(D, X) stop leaderboards (daily 416, 1h 220 instruments) Inconclusive · trust: medium
- Did the live-alert instruments fall from 69% to 55% win rate? Inconclusive · trust: low
- Audit of an earlier 64-ticker watchlist (KEEP / REMOVE) Inconclusive · trust: low
- Per-instrument pattern stats for the app (125-143 'masters') Inconclusive · trust: low
- Does the strategy work on crypto and commodities? No saved results · trust: low
- Equities vs forex vs crypto vs commodities No saved results · trust: low
- Broad 50-stock S&P sample across all 11 sectors No saved results · trust: low
- Live forward test of the 90 alerts Inconclusive · trust: low
- Is the edge spread across many stocks, or carried by a few? Supported · trust: high
- How much does survivorship flatter harmonic results and the edge over random? Supported · trust: medium
Market regime (4)
- Does the strategy do better when the market is falling (SPY below its 200-day MA)? Inconclusive · trust: medium
- Does the falling-market boost hold on 1h, 4h and daily over 10 years? Inconclusive · trust: medium
- Is the edge stable year by year? Inconclusive · trust: high
- Is there a hidden state that makes harmonics work at some times and not others? Not supported · trust: medium
Sizing (3)
- What would a 1%-risk paper account have done? Inconclusive · trust: low
- Is 5% of the account per trade survivable? Not supported · trust: medium
- Compounding $10,000 over 15 years: 1.5R vs 2R vs S&P buy & hold No saved results · trust: low
Machine learning (3)
- Can a gradient-boosted model predict which 15m setups win? Not supported · trust: high
- Can a neural network do better than the tree model? Not supported · trust: medium
- Can each instrument get its own best filter? Not supported · trust: high
Outside indicators & papers (6)
- Is the 'Harmonic Auto-Validator [GBB]' indicator's 89-98% win rate real? Not supported · trust: low
- Does the VP-MACD paper (SSRN 7545710) replicate, and does it beat buy & hold? Not supported · trust: medium
- Does the 'Hybrid AI trading system' paper (SSRN 6139568) reproduce? Not supported · trust: medium
- Is a Volumatic VIDYA [BigBeluga] trend flip a good entry? Not supported · trust: medium
- Backtesting the Wyckoff [theUltimator5] indicator as a strategy No saved results · trust: low
- Flags and pennants breakout strategy No saved results · trust: low
Education and research only. Not financial advice. Backtests use today's index members (survivorship bias).