Wyckoff Reality Check — Methodology & Locked Rules
Every rule behind the Wyckoff Reality Check was locked before the first number was seen (pre-registration) and has not changed since. This page states the detection rule, the measurement design and its limits.
- Level
- Confirmed 6/6 pivot low that is also the low of its 20-bar range
- Spring
- Undercut + close back above, at the bottom of the range, ≤ 3 prior breaks, entry at close
- Measurement
- Δ = forward log return (net 0.20% round-trip) − unconditional base rate; horizons 5/10/20 days
- Guards
- 30-event sample gate, placebo arm (RSI-14 crosses 30), Deflated Sharpe with 12 trials
- Look-ahead
- Single forward pass; prefix-invariant by construction, proven in tests
The measurement always subtracts the unconditional base rate of the same pair over the same window — without that difference, any post-pattern return mostly measures that crypto went up. Costs hit only the conditional arm, which is conservative against the spring. Results under 30 events are labelled anecdotes; a pattern that cannot beat a deliberately trivial RSI placebo carries no information of its own.
What this is not: a live signal, a recommendation, or a refutation of Wyckoff theory as a whole. One concrete, mechanical rule was tested on daily candles across 10 survivorship-clean pairs. Any different rule would need to be specified and measured the same way — before seeing the result, not after.
Wyckoff Reality Check — Methodology
All rules were locked BEFORE the first number was seen (pre-registration) and have not changed since. The tool shows exactly the tested rule — otherwise it would show something the finding does not cover.
1. The detection rule (locked)
- Pivot low: a bar whose low is strictly below the 6 bars to its left and not above the 6 bars to its right. Confirmed only 6 bars later — before that, the level does not exist.
- Valid level: the pivot low must also be the low of its 20-bar neighbourhood.
- Spring: a bar undercuts the level and closes back above it, sits at the bottom of its 20-bar range, and the level had at most 3 prior breaks. Each level fires at most once.
- Entry = close of the trigger bar. Never the low — you only know the low in hindsight.
- Volume split: V1 = trigger bar with ≥ 1.5× average volume (“absorption”), V2 = ≤ 0.7× (“no supply”). V0 is the pure structure without a volume condition.
2. The measurement
- Horizons: 5 / 10 / 20 trading days of forward log return from entry.
- Costs: 0.20% round-trip, subtracted ONLY from the conditional arm. The base rate stays gross — conservative against the spring.
- Base rate: the unconditional mean forward return over ALL bars of the same window. Δ = conditional − unconditional; without that difference you only measure that crypto went up.
- Gates: below 30 events a result is an anecdote and is labelled as such. The study additionally required ≥ 12 distinct calendar months and no single month holding > 50% of events.
- Placebo arm (mandatory): RSI(14) crossing above 30 — a deliberately trivial trigger. A pattern that cannot beat its placebo carries no information of its own.
- Multiple testing: Deflated Sharpe Ratio with 12 trials counted (3 variants × horizons + placebo).
3. Look-ahead freedom
Detection runs in a single forward pass: at every bar t, only data up to t is used. That is not a claim but built by construction and proven in tests — the same series, truncated at 25/50/75%, yields identical springs for the overlapping part (prefix invariance). The last, still-forming daily candle is always dropped.
4. Study universe
The 10 pairs were selected as of their 2020 standing (survivorship-clean: pairs delisted since — MATIC, EOS — stayed in the study), data from each pair's Binance listing through 2026-06-30: BTCUSDT, ETHUSDT, MATICUSDT, BNBUSDT, TRXUSDT, XRPUSDT, EOSUSDT, LTCUSDT, BCHUSDT, VETUSDT. The tool additionally applies the same rule to pairs outside the universe; the surface labels that.
5. What this is not
Not a live signal, not a recommendation, not a refutation of “Wyckoff theory”. What was tested is one concrete, mechanical rule on daily candles. If you mean a different rule, it needs to be specified and measured the same way — before, not after.