Stop guessing. Start building.

I fixed the rainbow moving average. It got noisier. The classic rainbow is five moving averages spaced 10, 20, 30, 40, 50. On a ratio scale that looks wrong: 10 to 20 doubles the length, 40 to 50 adds a quarter. The top two lines sit just 0.42 ATR apart and their daily changes correlate at 0.90. They move as one. So I respaced it by a constant ratio: 10, 15, 22, 33, 50. Even steps all the way up. I expected fewer false signals. The data said the opposite. A false flip here is a full alignment, all five lines in order, that dies inside 5 bars. The respaced stack had 3.03 a year against 2.37 for the classic one, more in 10 of 11 markets. Across 15 matched pairs and 11 futures, ratio spacing had more false flips in 138 of 165 comparisons. Look at where the room came from. The crowded top pair only ended 5.7% of alignments. The bottom pair, 10 vs 20, ended 74.5%. Ratio spacing opened the top and squeezed that bottom gap from 0.66 to 0.38 ATR, right where the stack breaks. Fixing the part that looks wrong can crowd the part that does the work. If you respace an indicator to make it look right, check which part breaks first. Full study: https://statoasis.com/overfit/research/rainbow-moving-average

RSI(2) below 10 still beats drift in every period I measured. That is exactly what makes it dangerous. It is the most famous buy-the-dip rule there is. Over 33.5 years of SPY it still looks fine: 440 trades, a profit factor of 1.61, a Strong verdict. That whole-window number is hiding something. Cut the same history into three periods with the same number of bars. Measure the entry in each one against that period's own drift, which is what buying at any day's open and selling four days later paid. 1993-02 to 2004-04: +0.549% a trade over drift, on 158 trades. 2004-04 to 2015-06: +0.198%, on 145 trades. 2015-06 to 2026-08: +0.096%, on 137 trades. Read the trade counts first. The rule never stopped firing. What changed is what each signal is worth: less than a fifth of what it paid in the first period. That is decay, not death. And nothing in the whole-window numbers shows it. RSI(2) is not alone. Of the 40 dip entries I tested, 11 kept less than half their early edge, 10 changed sign in at least one period and 3 fired too rarely to judge. 16 held up. Not everything is fading, either. PPO went +0.008%, +0.172%, +0.398%. Drop it on data ending in 2004, when it was barely above drift, and you would have missed the two periods where it paid most. So cut your own history into three and measure again before you trust the average. One honest limit: these are frictionless backtests on SPY, so +0.096% is before costs. The stability verdict for all 40 entries: https://statoasis.com/overfit/research/mean-reversion-buy-the-dip-ranked

Five dip entries can be one trade. Two different formulas can fire on the exact same days. Different names, different math, same trades. You think you have diversified. You haven't. So for the second edition of 36 Ways to Buy the Dip, I checked all 780 pairs of entries for firing on the same days. Some entries are the same trade in different clothes. The book shows you which ones add nothing. That check sits beside two others. First, every entry had to beat a random entry held just as long. Second, every entry had to hold up in three equal periods of SPY history. Only 16 of the 40 did. You can read the full RSI(2) entry free on the page, so you see the format before you buy. The $79 launch price ends tomorrow, Thursday 5 November, at midnight New York time. After that it is $147. https://statoasis.com/books/36-ways-to-buy-the-dip