<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
	<channel>
		<title>Overfit</title>
		<link>https://overfit.info</link>
		<description>A public archive of trading strategies that did not work, with the receipts: rules written down before the test, every try counted, results across markets, and an honest verdict. Research and education only, not investment advice.</description>
		<atom:link href="https://overfit.info/rss.xml" rel="self" type="application/rss+xml" />
		<item>
			<title>The Benner cycle (1875): beautiful chart, untestable</title>
			<link>https://overfit.info/posts/benner-cycle</link>
			<guid isPermaLink="true">https://overfit.info/posts/benner-cycle</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>A famous chart of &quot;good times&quot; and &quot;hard times&quot; years. We have about 21 years of daily data, so we cannot test it statistically.</description>
			<category>untestable</category>
			<category>esoteric</category>
		</item>
		<item>
			<title>Wait for the stars to align: 24 setups, 0 survivors</title>
			<link>https://overfit.info/posts/confluence-24-setups</link>
			<guid isPermaLink="true">https://overfit.info/posts/confluence-24-setups</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>Sit in T-bills until several conditions line up, then hold for 1, 3 or 6 months. None survived once we counted the tries.</description>
			<category>overfit</category>
			<category>setups</category>
		</item>
		<item>
			<title>1,192 indicator combinations: the winner was no better than luck</title>
			<link>https://overfit.info/posts/grid-1192-combinations</link>
			<guid isPermaLink="true">https://overfit.info/posts/grid-1192-combinations</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>We searched every 1-to-3 indicator combination from 12 signals. The best looked fine until we ran the same search on scrambled data.</description>
			<category>overfit</category>
			<category>models</category>
		</item>
		<item>
			<title>144 signal tests, two winners, no edge</title>
			<link>https://overfit.info/posts/ic-screen-144-tests</link>
			<guid isPermaLink="true">https://overfit.info/posts/ic-screen-144-tests</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>We screened 24 popular indicators at three horizons. Two passed a false-discovery check. Neither beat what pure chance produces across the whole family.</description>
			<category>overfit</category>
			<category>diagnostics</category>
		</item>
		<item>
			<title>Moon phases: new moon beat full moon in one window and lost in the other</title>
			<link>https://overfit.info/posts/lunar-cycle</link>
			<guid isPermaLink="true">https://overfit.info/posts/lunar-cycle</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>A sign test on lunar-phase returns. It failed the 2017–2021 window and parked the whole astro-timing family.</description>
			<category>edge-decayed</category>
			<category>esoteric</category>
		</item>
		<item>
			<title>Pre-Fed-announcement drift: real, but it happens overnight</title>
			<link>https://overfit.info/posts/pre-fomc-drift</link>
			<guid isPermaLink="true">https://overfit.info/posts/pre-fomc-drift</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>The premium before Fed meetings is documented. In our data the gain arrived overnight, before a next-morning fill could catch it.</description>
			<category>not-tradable</category>
			<category>calendar</category>
		</item>
		<item>
			<title>Indicator families and pairs: best was good, not distinguishable from luck</title>
			<link>https://overfit.info/posts/signal-families-28-books</link>
			<guid isPermaLink="true">https://overfit.info/posts/signal-families-28-books</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>Momentum plus risk led 28 family and pair combinations. It cut the worst drop in half and still could not clear the overfitting checks.</description>
			<category>overfit</category>
			<category>models</category>
		</item>
		<item>
			<title>The plain 200-day rule: halves the drops, gives up the growth</title>
			<link>https://overfit.info/posts/sma200-overlay</link>
			<guid isPermaLink="true">https://overfit.info/posts/sma200-overlay</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>One simple, untuned rule written down in advance. It passed one validation window and failed the other.</description>
			<category>not-worth-it</category>
			<category>trend</category>
		</item>
		<item>
			<title>A take-or-skip judge on top of our setups: it took nearly everything</title>
			<link>https://overfit.info/posts/take-or-skip-gate</link>
			<guid isPermaLink="true">https://overfit.info/posts/take-or-skip-gate</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>We asked a decision gate to approve or reject signals from our earlier setups. It either approved nearly every one or its picks were no better than its skips.</description>
			<category>not-worth-it</category>
			<category>models</category>
		</item>
		<item>
			<title>Turn-of-month plus Fed days: more trades, same disappointment</title>
			<link>https://overfit.info/posts/tom-plus-fomc</link>
			<guid isPermaLink="true">https://overfit.info/posts/tom-plus-fomc</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>Combining the two calendar effects did not fix either.</description>
			<category>not-worth-it</category>
			<category>calendar</category>
		</item>
		<item>
			<title>32 trend variants: the timing was real, picking the winner was not</title>
			<link>https://overfit.info/posts/trend-regime-32-variants</link>
			<guid isPermaLink="true">https://overfit.info/posts/trend-regime-32-variants</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>The best trend variant looked superb on 2006–2016. The overfitting checks said it was mostly luck among near-identical cousins.</description>
			<category>overfit</category>
			<category>trend</category>
		</item>
		<item>
			<title>Turn-of-the-month: invested 12% of the time, earned 4% a year</title>
			<link>https://overfit.info/posts/turn-of-month</link>
			<guid isPermaLink="true">https://overfit.info/posts/turn-of-month</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>The turn of the month has a documented return premium. As a standalone rule it left money in cash almost all the time.</description>
			<category>not-worth-it</category>
			<category>calendar</category>
		</item>
		<item>
			<title>The drawdown brake: passed on paper, not proven with money</title>
			<link>https://overfit.info/posts/v-overlay-drawdown-brake</link>
			<guid isPermaLink="true">https://overfit.info/posts/v-overlay-drawdown-brake</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>A 200-day trend filter with a 10% volatility target cut the worst drops in 6 of 7 new markets. It also gave up a lot of return. It has not been run with real money.</description>
			<category>passed-unproven-live</category>
			<category>trend</category>
			<category>volatility</category>
			<category>cross-market</category>
		</item>
		<item>
			<title>Volatility-targeting plain SPY, VTI and VUG: close, not enough</title>
			<link>https://overfit.info/posts/vol-target-core-etfs</link>
			<guid isPermaLink="true">https://overfit.info/posts/vol-target-core-etfs</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>Scale an equal-weight mix of three broad funds to a 10% volatility target. It fell short of the pre-set Sharpe bar: a 0.035 shortfall where 0.02 was allowed.</description>
			<category>not-worth-it</category>
			<category>volatility</category>
		</item>
		<item>
			<title>Machine learning on 24 indicators: no better than shuffled data</title>
			<link>https://overfit.info/posts/walk-forward-models</link>
			<guid isPermaLink="true">https://overfit.info/posts/walk-forward-models</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>Regularized regressions and a small tree model, retrained every year, did no better than the same models trained on scrambled answers.</description>
			<category>overfit</category>
			<category>models</category>
		</item>
		<item>
			<title>Weekly reversal: great on 2005–2016, nothing on 2017–2021</title>
			<link>https://overfit.info/posts/weekly-reversal</link>
			<guid isPermaLink="true">https://overfit.info/posts/weekly-reversal</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>Buy the most oversold ETFs and hold a few weeks. The training years loved it. The pre-registered validation years did not.</description>
			<category>edge-decayed</category>
			<category>reversal</category>
		</item>
		<item>
			<title>Yield-curve alarm: right in 2017–2021, almost all cash in 2022–2023</title>
			<link>https://overfit.info/posts/yield-curve-gate</link>
			<guid isPermaLink="true">https://overfit.info/posts/yield-curve-gate</guid>
			<pubDate>Fri, 20 Nov 2026 00:00:00 GMT</pubDate>
			<description>Go to cash when the yield curve inverts. It helped in one window, then sat out most of the next.</description>
			<category>edge-decayed</category>
			<category>macro</category>
		</item>
	</channel>
</rss>
