Blog
Data-driven insights on seasonal trading patterns, cycle timing, and market analysis.
Jake Bernstein’s Swing Methods on the Futures in His Weekly Report: 21 Backtests on 29 Markets
We ran 21 daily and weekly swing configurations inspired by Jake Bernstein on 29 of the 39 markets his weekly report covers, 2002 to October 2026, after illustrative futures costs. 3 of 21 finished with a positive mean per trade, all of them weekly, with bitcoin contributing much of their summed equal-notional net returns. Here are all the results by market, period and cost.
Swing-Trading Backtests Inspired by Jake Bernstein: 21 Research Configurations
Simulations on 342 stocks listed in the US with next-open execution, lagged liquidity screens, independent date windows and explicit limits. Historical trade averages do not establish alpha.
Do George Bayer’s Astrological Trading Rules Work? We Tested 55 of Them on 764 Markets
In 1940 the wheat trader George Bayer published rules that turn Mercury’s speed, Venus’s declination, Saturn’s latitude and dozens of other planetary motions into dates when markets must turn. We computed mechanical readings of 55 of his 61 rules, matched 313 of the 333 example dates he printed, and tested them on 764 markets from the Dow in 1896 to wheat futures in 2026. His dates were swing highs or lows 38.6% of the time, the same as random dates; one of 56 statistical tests cleared p < 0.05 where 2.8 would be expected by chance, and none survived the correction for 56 tests. Wheat and his own years, 1932 to 1940, showed nothing either. The reason his examples look so good: 97.4% of all days have one of his rules within a day.
Do Harmonic Patterns Work? We Tested 896,068 of Them Against Fake Fibonacci Ratios
A TradingView chart’s Shark target printed to the tick on the October 14, 2026 CPI bar. We detected every Gartley, Bat, Butterfly, Crab, Shark and Cypher on 10 years of 5-minute bars for 544 US stocks and ETFs, with no hindsight. They reached their first target 45.7% of the time, against 46.1% expected from their own target and stop distances. Patterns built from deliberately wrong ratios did just as well, a random walk looked the same, and stricter definitions from Carney and TradingView changed nothing material.
TradingView ICT Master Suite Best Settings
We searched 398 combinations of ICT Master Suite’s TradingView settings on Nasdaq futures, 2016 to September 2026, and checked the best one on later data it never saw. It made money in both periods, with a profit factor of 1.53 and then 1.27, but it needs two filters the script doesn’t have and per trade it did no better than random dates. With only the script’s own inputs it about broke even, the defaults lost on recent data, and at the script’s default commission every setting loses when each trade risks the same amount. Here are the results, the profit factors and the exact settings.
Price Cycle Analysis, Explained: How the Cycle Screener Fits, Tests and Projects 688 Markets
A walkthrough of the Price Cycle Screener and Analyser using the free E-mini S&P 500 model: how cycles are found in daily prices, turned into trades, frozen at a cutoff and tested on the next nine months after costs; why the screener shows the best of 5,000 settings and what that does to the numbers; and how the new Cycle aligned comparison checks a seasonal window against the price cycle.
ICT Calls It a “Two-Stage Delivery.” We Tested It on NFP, CPI, PPI and FOMC. It Went 0 for 4.
ICT’s rule for news mornings: let the number print, wait 15 to 20 minutes, then trade back into the range the move created — he calls it a “two-stage delivery” and says a similar shape, not always, shows up on CPI, PPI mornings and the Fed’s afternoon moves too. We built the trade itself for NFP and CPI/PPI, tested against a fairly-built comparison trade, and measured the pattern itself for FOMC against an ordinary afternoon. Across 86 fired NFP trades since 2016, the wait-and-fade averaged about a fifth of a risk unit lost per trade (−0.222R) — the closest of the four claims to a real effect, but its plausible range still touches a small profit. CPI and PPI disagree in sign when tested separately, and FOMC afternoons show no detectable reversal edge over an ordinary Wednesday. None of the four claims survives the stricter bar required when several are tested together.
The Same 54 Bernstein Configurations on Gold: What Survived the Move From Nasdaq
Every intraday configuration from our NQ study, re-run on spot gold at COMEX contract economics: which ones stay positive, which flip sign, and how little the two markets agree.
ICT Says Half the Opening Gap Fills by 10 A.M. 70% of the Time. We Tested the Wrong Side Too.
ICT’s “How New Students Should Start” series says the midpoint of the opening range gap trades by 10:00 ET about 70% of the time. On 2,636 NQ futures sessions it did on 57.7%, and a level the same distance from the open in the opposite direction on 56.1%. Our mechanical versions of his three daily pillars showed no detectable edge before costs. Of ten claims under a Holm correction, one held: NFP Friday mornings are about 18% wider.
54 NQ Backtests Inspired by Jake Bernstein: An Independent Study
54 intraday research configurations based on Bernstein’s course and books, tested on NQ futures. Compare costs, weekday selection and an additional 240-setting dual-moving-average grid.
We Backtested ICT Master Suite. Its 17% Came From Fifteen Trades.
We ran all ten ICT Master Suite configurations on Nasdaq futures. The best, Silver Bullet long only, made +17.36% in fourteen weeks at a 66.20% win rate over 71 trades, and all of it came from 15 New York PM trades. At the script’s default 0.02%-per-side setting, commission took 49% of the profit before costs. We rebuilt the model in Python and matched 43 of its entries to the same five-minute bar. Run long only over 1,877 trades and 8.5 years at that commission rate, it loses 0.130R per trade with no profitable year; at 2 points of cost it is close to breakeven.
We Watched Trader Mane’s Entire ICT Course. Then We Backtested It.
We froze the choices left open by Trader Mane’s 20-video ICT course into four causal strategies and six controls, then tested them on 2.6 million Nasdaq minute bars and replicated on GBPUSD. The strict variants produced 0–3 trades each; the higher-sample controls lost after costs. None passed the registered evidence gates.
We Backtested Four ICT Setups Against Random Trades. None Held Up.
All four ICT setups we backtested (fair value gaps, Power of Three, inversion FVGs and breaker blocks) made money before costs on 8.5 years of Nasdaq data, and so did random trades. Against the same trades on random dates, three could not be told apart. Breaker blocks beat them on Nasdaq, then failed to replicate on GBPUSD. A 15-minute FVG variant came out borderline and lost money at one point of cost.
We Read All 620 ICT Videos. The “Codeable” Silver Bullet Wasn’t Actually Mechanical.
The Silver Bullet looked codeable until we audited the primary lecture: direction comes from a preselected draw on liquidity. ICT teaches ways to judge that draw, but the Silver Bullet material gives no universal mechanical tie-breaker when several eligible draws coexist. We tested all seven transcript-listed draw categories across 2.6 million Nasdaq 1-minute bars using causal fills and two points of cost. No literal draw produced a statistically defensible edge; the least-bad version still lost 0.021R per trade.
The Fed Crushed Stocks Wednesday. By Friday It Never Happened. We Tested All 24 Since 1952.
On July 29, 2026 the S&P 500 fell 1.52% on the Fed decision and had erased the entire loss by Friday’s close. That shape has occurred 24 times since 1952, and the Monday after closed higher 14 of them — 58.3%, the kind of stat that gets quoted all weekend. Then we ran the control: any Friday closing up 1.5%+ is followed by a 57.2% Monday. The edge is end-of-week momentum, not the Fed. Interactive table with a chart for every precedent.
ICT’s “Chain of Custody” Lecture, Turned Into Code
ICT’s July 2026 lectures use two graded ranges at the same time: a projection from an imbalance candle to a chosen liquidity draw, and a printed daily inefficiency graded from its own high to its own low. We encoded both. The levels are deterministic after their inputs are fixed, but selecting the projected target — and deciding which daily zone remains active — still requires a stated rule.
We Coded All Five Setups From a Viral ICT Strategy Guide. One Survived (Barely).
A popular TradingView guide teaches five ICT setups on Bitcoin — fair value gaps, Power of Three, inversion FVGs, breaker blocks, the Silver Bullet — with plenty of annotated charts and zero numbers. We mechanized all five and ran them over 10,000+ trades on BTC perpetuals and 1-minute NASDAQ futures with conservative fills and real costs. On Bitcoin, stop geometry hands 0.3–2.4R per trade to fees; on NQ, three setups land at breakeven and inversion FVGs lose at scale. The one survivor — 15-minute FVG retrace with a displacement filter — netted +0.155R over 249 trades, every one listed for review.
Mercury Combust AND Retrograde: We Tested the Double Condition on 76 Years of Data
If retrograde scrambles markets and combustion burns up Mercury's signal, the stretch where both hold should be the danger zone — and the 2008 crash week and the COVID crash both started inside one. We computed the combust-retrograde window of all 242 retrogrades since 1950 (it exists in every single one), tested drops after entering it and volatility inside it, with a clickable chart for every window. Inside vs. outside: daily range 1.19% vs 1.17%, up days 53.2% vs 53.1%. The market cannot tell the difference.
Does the Market Drop at Least 2% When Mercury Goes Retrograde? We Tested All 242 Since 1950
The claim: expect a ~2% drop in the days after Mercury stations retrograde — and again at the midpoint. The highlight reel is real: Mercury stationed the Friday before Black Monday 1987 and three days before the COVID crash. So we tested all 242 stations and midpoints since 1950 at four drop thresholds, with a clickable chart for every event. The base rate nobody mentions: 2% five-day drops follow 18.2% of ALL trading days. The stations manage 17.8%.
The Venus 225-Day Square-Out: We Projected One Venus Orbit From Every Major S&P Pivot Since 1950
A Gann-style call from a trading clinic: July 3, 2026 sat exactly one Venus orbit — 225 days — after the November 20 low, so the cycle had "squared out" and fireworks would start Monday. The arithmetic is perfect. We projected all 519 major S&P 500 pivots since 1950 forward 225 days, tested swing pivots and tradeable reversals against their base rates, and swept every offset from 30 to 400 days — Venus ranks #98 of 371. The market answered too: a five-week high a week later.
New Moon and Mercury Combust on the Same Day: We Found All 20 Since 1950
An astro-trading clinic asked for research: how often does the New Moon land on the same day as a Mercury-Sun conjunction, and are those large-range days? We computed every occurrence since 1950 — 20 events, about one every four years — and checked the S&P 500 on each, with an interactive chart for every date. The rarity is real. The volatility is not.
Does the Market Turn at the Midpoint of Mercury Retrograde? We Tested All 241 Since 1950
A trading-community claim: the exact midpoint of Mercury retrograde — the night of July 12 for the current one — marks a market direction change. The astronomy checks out; it is the inferior conjunction. So we ran the full battery on all 241 midpoints since 1950: swing pivots, tradeable 1.5% reversals, a walk-forward hot/cold cycle filter, and every astrological slice from combustion to Mercury-in-Leo — with an interactive table where you can open the chart for every single one. Nothing beats the random-day base rate.
Does the Market Really Reverse on Perihelion and Aphelion? We Checked 76 Years of Data
A claim from astro-trading circles: the S&P 500 "always" reverses when Earth is closest to or farthest from the Sun. We computed the exact dates for 1950-2026 and ran the base-rate test nobody runs — with an interactive table of all 153 events where you can open the chart for every single date. The streak is real — 10 of 11 recent perihelions "hit" — and so is the reason it means nothing.
We Coded Two “70% Win Rate” ICT Session Strategies. Neither Survived.
The follow-up to our fair value gap study: a NY 8:30 range-reversal on NASDAQ futures and a London-bias order-block setup on GBPUSD, coded exactly as taught and run over 523 trades with conservative fills. Win rates came out at 18–29% against 29–33% breakevens — the advertised 70% lives in the discretion, not the rules.
We Backtested ICT’s Fair Value Gap Rules on 53 Years of Dollar Index Data
A tick-perfect Dollar Index reversal call in June 2026 sent us to the code editor. We mechanized the fair value gap rules — volume imbalances, the validity filter, consequent encroachment — and ran 6,035 gaps and 3,986 trades since 1973. Win rate: 38%. Expectancy: +0.29R. The filter and the midpoint exit both earn their keep.
The Truth About 100% Win-Rate Seasonal Patterns
We stress-tested 37,430 patterns with a perfect 100% record across 1.4 million trades back to 1981. An early bad day doesn't predict failure, the 100% record falls to ~57% out of sample, and only two things actually hold up: the pattern's historical magnitude and the calendar month of entry.
"Sell in May" Is Wrong — But the Cycle Is Turning Down
For the last 10 years, buying the S&P 500 on May 20 and selling on June 9 has won 10 out of 10 times with an average return of +2.5%. But the equity curve cycle that has guided this pattern since 1950 is now rolling over.
The 60-Year Cycle: Trading the S&P 500 With Swing Pivots From 1960
Swing highs and lows from 60 years ago mapped onto today's S&P 500. Long-only, the strategy produces a 70.8% win rate, 4.35 profit factor, and +372% return over 16 years — but the current regime is COLD.
How to Use Equity Curve Regime Filtering to Avoid Cold Streaks
A seasonal pattern with an 80% win rate can still lose money for years. Equity curve regime filtering tells you whether a pattern is working right now — turning -8% into +128% with the same patterns.
What Is Seasonal Trading? A Data-Driven Guide
How recurring price patterns work, how to evaluate them with win rates and efficiency scores, and why some of the most well-known seasonal tendencies have persisted for decades.
ICT First Hour Dealing Range + FVG: Coin Flip or Edge?
We backtested ICT's claim: when a candle body breaks the first hour's dealing range, enter on the first presented FVG. 12 years of 1-minute DJI data (1.2M bars, 3,180 trading days) analyzed. The "trending day" qualifier is doing all the work.
How to Screen Stocks for Seasonal Patterns
Find high-probability seasonal setups in seconds — not hours. A data-driven walkthrough of systematic seasonal screening with a real DHL.DE example: 10 trades, 10 winners, 100% win rate.
ICT Turtle Soup + FVG Carryover: Does It Actually Work?
We backtested ICT's claim that carrying a Fair Value Gap from a turtle soup during the 2-hour launch period to the next day predicts the high or low of the day. 6 months of 1-minute NQ futures data analyzed.