The Overfit archive
Trading ideas are cheap. Evidence isn’t.
We put popular trading ideas through the data, and publish what works, what fails, and what only looked good on the right chart.
48 articles · 46 under the research standard · Every result published, including the ones that failed
#153Strategies, Backtested
Rainbow Moving Average Settings: Spacing Is a Coin Flip in 165 Tests
Oct 29, 2026 · 10 min read
The classic five-line rainbow is spaced evenly (10, 20, 30, 40, 50), the way Widner's original stepped its lag. Spacing it by a constant ratio instead (10, 15, 22, 33, 50) looks like the fix, and across 11 futures since 2005 it is not: it added false alignment flips in 138 of 165 matched comparisons and won on return to drawdown in 87 of 165, a coin flip. The fastest line is what places a stack in the stable region.
#146Backtesting & Validation
Does the COT Report Actually Work? 4,889 Backtests on Larry Williams' COT Index
Oct 22, 2026 · 13 min read
Across 4,889 backtests of Larry Williams' COT index on E-mini S&P 500 futures, 92.6% of filter-on variants beat the unfiltered baseline's $586.67 average trade. Not one of the 4,888 beat its $238,187.50 net profit or its 0.51 CAR/MaxDD.
#145Backtesting & Validation
S&P 500 Opening Gaps: Direction Is Noise, Size Is the Signal
Oct 8, 2026 · 11 min read
Across 8,397 S&P 500 opening gaps since 1993, direction is almost noise, with up-gaps and down-gaps both drifting about +0.2% over five days, while size carries the signal: gaps over 2% average +0.368% the next day, 94-98% of gaps fill, and every one of the top ten variants by CAR/MaxDD is long while the median short variant loses under all three lenses.
#144Market Behavior & Concepts
S&P 500 Big Range Days: A Big Day Predicts Volatility, Not Direction
Oct 1, 2026 · 11 min read
"Big range day" means three different things depending on whether you measure in points, percent of price, or ATR, and the rulers disagree about which days even count: 549 of the 635 six-point days landed in the 2020s, as SPY went from $44.34 to $741.75.
#143Strategies, Backtested
RSI vs Stochastic vs Williams %R: 33,792 Backtests
Sep 24, 2026 · 12 min read
33,792 backtests across RSI, Stochastic and Williams %R on four index ETFs: the setup matters more than the indicator. 89.7% of long-side variants made money against 8.8% of short-side, and buying dips beat chasing breakouts 55.7% to 42.9%.
#142Strategies, Backtested
The Better-RSI Showdown: We Tested 4 RSI Upgrades on SPY, QQQ, IWM, and DIA
Sep 10, 2026 · 10 min read
1,856 backtests across four RSI families on SPY, QQQ, IWM and DIA: Connors RSI and Z-Score RSI modestly beat plain RSI on median risk-adjusted return, Laguerre RSI was the worst of the four despite its lag-free marketing, and the simplified three-condition Triple RSI tested here does not reproduce the popular win-rate claim.
#141Strategies, Backtested
I Backtested ICT / Smart Money Concepts — What Survives
Sep 3, 2026 · 10 min read
The four core ICT / Smart Money Concepts entries — order blocks, fair value gaps, liquidity sweeps and Optimal Trade Entry — codified into mechanical rules and run against three textbook entries and a coin flip across four markets. None showed a statistically significant 5- or 10-day forward-return edge on SPY.
#140Tools, Software & Tech Stack
Can AI Build a Profitable Trading Strategy? I Backtested 5 LLM-Generated Rules on SPY to Find Out
Aug 22, 2026 · 11 min read
Two experiments on SPY: a 2,304-variant optimizer sweep standing in for what AI does when it hunts for "the best strategy," and 5 AI-generated rules implemented verbatim. The optimizer's top-25 risk-adjusted score collapsed 85% out-of-sample, and 0 of 2,068 eligible strategies beat buy-and-hold.
#139Backtesting & Validation
Coin Flip Trading Strategy: 40,000 Random Strategies Reveal Market Bias
Oct 11, 2025 · 8 min read
I let a coin pick the trading days for 40,000 strategies on ES, SPY, natural gas and the yen, 2007 to 2026. 97.7% of random ES longs made money. Almost half reached a profit factor of 1.3 or more. But 82.6% of natural gas's fall was the cost of rolling the contract, and about a third of a real RSI(2) rule's average trade was drift a coin earned anyway.
#138Portfolio & Risk
Dollar Cost Averaging vs Lump Sum vs a 200-Day Moving Average: 273 S&P 500 Start Dates
Aug 17, 2025 · 7 min read
I put the same $120,000 into the S&P 500 three ways on 273 monthly start dates: all at once, spread monthly, and with a 200-day moving average switch. Lump sum ended richer than 12-month DCA in 76.9% of 10-year windows, and spreading it over 12 months barely changed the worst fall. The switch cut the median fall from 47.5% to 20.0% and never ended a decade below the money put in, but plain lump sum still ended with more in 81.0% of windows.
#137Strategies, Backtested
Keltner Channels vs Bollinger Bands: 116,640 Backtests (and the Third Band I Built)
Aug 2, 2025 · 12 min read
116,640 backtests put Bollinger Bands, Keltner Channels and Casey Bands through one identical harness. Measured the same way on both, the band you pick moves the share of profitable long variants by 2.8 percentage points and the side you trade moves it by 90.4. At each family's most durable SPY setting, paying 0.05% a side drops Bollinger below buy-and-hold on MAR while the other two stay ahead.
#136Strategies, Backtested
Larry Connors R3 Strategy: 8,067 Backtests on SPY and Index Futures
Jul 26, 2025 · 9 min read
Larry Connors' R3 rules still make money on SPY, ES and NQ from 2001 to 2026, but only about 2.8 times a year. Loosen the rules and you get far more trades, but on ES only 15.2% of loosened versions beat R3's average trade. ES has the same edge as SPY, just on a bigger contract.
#134Strategies, Backtested
Breakout Trading Strategy: Same Rules, 480 Backtests on Nasdaq, S&P 500 and Dow
Jun 20, 2025 · 7 min read
The same daily breakout trading strategy tested 480 ways on the Nasdaq, S&P 500 and Dow (QQQ, SPY, DIA, ES and NQ futures, 2007 to 2026, next-open fills): 441 variants made money, yet only 1 of 480 beat the average trade of 95 of its 100 random-entry runs with the same size and hold. The Nasdaq's bigger breakout dollars came with random Nasdaq entries earning 1.71x random S&P 500 entries.
#133Strategies, Backtested
CCI Trading Strategy: 672 Settings Tested on SPY and ES, and Where the Default Ranks
Jun 13, 2025 · 7 min read
The textbook CCI setting is CCI(20): buy below -100, short above +100. I tested it against 672 CCI settings on SPY since 1994 and ES futures since 2006. The default long rule beat 83.0% of reliable long settings on SPY but only 38.9% on ES. A faster CCI did better, the short side lost on both, and RSI(2) below 10 beat all but 6 of 223 reliable long CCI settings on SPY.
#132Strategies, Backtested
Donchian Channel Strategy: I Backtested '40 In, 20 Out' 1,188 Ways on 33 Years of SPY
Jun 6, 2025 · 8 min read
The classic 40-in/20-out channel breakout, re-tested 1,188 ways on 33.4 years of SPY: long-only still made 105.5% frictionless — but the famous 20-day-low exit lost to a plain 20-bar time exit, every unfiltered short entry length lost money under the channel-20 cover and three of the six drew down past 100%, and the edge has decayed every decade.
Browse all 48 articles
- Donchian Channel: What It Is, How to Trade It, and 900 Backtests on 15 Markets
- Rainbow Moving Average Settings: Spacing Is a Coin Flip in 165 Tests
- Does the COT Report Actually Work? 4,889 Backtests on Larry Williams' COT Index
- S&P 500 Opening Gaps: Direction Is Noise, Size Is the Signal
- S&P 500 Big Range Days: A Big Day Predicts Volatility, Not Direction
- RSI vs Stochastic vs Williams %R: 33,792 Backtests
- The Better-RSI Showdown: We Tested 4 RSI Upgrades on SPY, QQQ, IWM, and DIA
- I Backtested ICT / Smart Money Concepts — What Survives
- Can AI Build a Profitable Trading Strategy? I Backtested 5 LLM-Generated Rules on SPY to Find Out
- Coin Flip Trading Strategy: 40,000 Random Strategies Reveal Market Bias
- Dollar Cost Averaging vs Lump Sum vs a 200-Day Moving Average: 273 S&P 500 Start Dates
- Keltner Channels vs Bollinger Bands: 116,640 Backtests (and the Third Band I Built)
- Larry Connors R3 Strategy: 8,067 Backtests on SPY and Index Futures
- Breakout Trading Strategy: Same Rules, 480 Backtests on Nasdaq, S&P 500 and Dow
- CCI Trading Strategy: 672 Settings Tested on SPY and ES, and Where the Default Ranks
- Donchian Channel Strategy: I Backtested '40 In, 20 Out' 1,188 Ways on 33 Years of SPY
- StrategyQuant X Review: The No-Code Way to Build Trading Strategies
- Cumulative RSI Strategy vs RSI(2): 18,480 Backtests on the S&P 500
- Range Expansion Index (REI) vs RSI(2): 3,321 Trades on SPY, QQQ, DIA and ES
- The #1 Lesson Traders Learn Too Late: No Single System Will Save You
- Weekly Mean Reversion Strategy: 22,680 Backtests Say the Rule Works and the Volume Filter Does Not
- Volume Oscillator: 105 Settings Tested on ES, SPY and the Yen, and Whether Any Signal Holds Up
- VIX Explained: What a High VIX Means, From 35 Spikes Since 1990
- Internal Bar Strength (IBS) Strategy: 200 Backtests on Nasdaq and S&P 500 Futures
- Ultimate C%: A Mean Reversion Indicator That Trades Different Days From RSI(2), Tested on 32 Years of SPY and 20 of ES
- Robustness Testing: Why Most Traders Fail, and What 36,252 Backtests Say Actually Works
- Z-Score Mean Reversion Strategy: 2,400 Backtests on Futures and SPY, and the 74% Win Rate Is a Trap
- AI Trading Strategy Development: The Exact Prompts, Step by Step
- Tactical vs Strategic Asset Allocation: 18 Years of Backtests, After Trading Costs
- Market Regime: What It Is and What the S&P 500 Did Next in Each of 6 Regimes (32 Years)
- Is Trading a Skill or Luck? What 5,000 Random Rules and a Seeded Coin Flip Say
- Water Fasting for Traders: How to Boost Focus & Clear Brain Fog
- S&P 500 RSI Strategy: What 293,760 Backtests Found on SPY and ES
- CaseyC% Oscillator: A Mean Reversion Indicator That Trades Different Days From RSI(2), 32 Years of SPY and 20 of ES
- Random Walk Theory, Tested: Are Markets Random? 5 Tests on 8 Markets Against Their Own Shuffled Returns
- Cyclic RSI vs RSI: 11,640 Backtests on S&P 500 Futures and SPY
- OBV MACD vs. Traditional MACD: 9,216 Backtests Say You Are Asking the Wrong Question
- Do Volume and Volatility Filters Actually Improve RSI(2)? 15,552 Backtests Say Mostly No
- Ray Dalio's Holy Grail of Investing, Tested: Four Asset Classes Beat Nine Funds
- ADX Indicator Tested: 8,370 Days Say It Measures Size, Not Direction
- Larry Connors Double 7 Strategy: The Exact Rules, Tested 4,320 Ways on 33 Years of S&P 500 Data
- Awesome Oscillator Tested: 14,000 Backtests, and 5/34 Ranks 28th
- Retirement Bucket Strategy, Tested: 749 Start Dates vs a Plain Rebalanced Portfolio
- USDJPY Candlestick Chart Patterns: 13 Tested, None Beat Chance
- S&P 500 Drawdowns Since 1871: Every Decline, How Long They Lasted, and What Actually Recovered
- Monte Carlo for Traders: I Checked Whether the Simulation Was Right. It Wasn't.
- Toby Crabel's Thrust Patterns: 17,424 Backtests. Random Entries Matched the Up Thrust on Half the Markets
- VIX Trading Strategy Tested: 2,520 Backtests Say the Roll Pays, Not the Timing