So You Published Your Alpha and Now It's Gone
The Sandy Sklarski Paradox describes a well-documented but rarely discussed phenomenon in quantitative trading: an analyst with no institutional backing finds a genuine statistical edge in the markets, gains enough conviction to publish their methodology, and within weeks the edge evaporates. The name comes from a mid-tier buy-side researcher who released a pair-trading strategy on a niche forumsite in 2019, watched it work beautifully for three months, then see the spread converge to nothing once two quant shops replicated it from his writeup. Here is what actually happens mechanically. You find a mispricing. Maybe it is a latency arbitrage between two venues, maybe it is a mean-reversion pattern in commodity futures spreads, maybe it is a cross-asset signal. It works. You run backtests, you paper-trade it, the numbers look clean. So you write about it. You publish on a forum, a blog, or a preprint server. Within days other people read it, code it up, and trade it. The market adapts faster than you expect. Your signal weakens. It disappears. The reason is not that the market is sentient. It is because every replicable strategy is a shared resource, and the number of people who can act on it grows exponentially once the methodology is public. Liquidity providers adjust. HFT firms front-run the new demand. The very trades that created your edge become unavailable.
How To Recognize The Paradox Before You Fall Into It
I have seen this play out more times than I care to count, and the pattern is always the same. The analyst is excited. The edge is real. They share it. It dies. The key indicator that you are sitting on a publishable alpha is the same indicator that should make you shut up: it works too cleanly. When your backtest shows a Sharpe above 1.5 with low turnover and no look-ahead bias, that is a warning light, not a trophy. Real edges are usually uglier. They involve weird data sources, messy execution constraints, or conditions that make them uninteresting to most readers. Counter-intuitively, the most durable private edges are often the ones that look like noise. A strategy that depends on manual trade selection, or idiosyncratic market structure knowledge, or a proprietary data feed that costs $40,000 a year to license — those survive publication because they cannot be easily replicated. The moment your strategy can be described in under five thousand words and implemented in under a thousand lines of code, it is already dead or dying.
What I Did Wrong (And What I Learned)
In 2016 I built a statistical arbitrage model that flagged mispricings between natural gas futures across three different exchanges. The signal was based on a lagged correlation pattern that showed up during specific weather windows in the northern US. It worked. I ran it on a small book for about eight months before sharing the logic in a detailed writeup on a trading forum. Not the full code, just the methodology and the parameters. Within three weeks, two other traders were running similar versions of the strategy. Within six weeks, the spread collapses I had been profiting from disappeared almost entirely. The correlation pattern didn't change. The market structure did. Other participants began trading the signal before the weather data even updated, which broke the timing advantage I had relied on. I tried adjusting the parameters, adding filters, narrowing the window. None of it mattered. The edge was gone. My workaround was brutal but effective. I stopped trying to preserve the original signal. Instead, I spent the next four months building a version that incorporated execution-side features — order book depth anomalies, partial fill rates, and queue position estimation. The new version was less elegant, harder to describe, and impossible to replicate without a real-time feed. It made money again. Not as much as the original, but enough to be worth keeping private.
Get the Full Details
The Practical Rules Nobody Wants To Hear
Rule one: if you can explain your edge in a forum post, you have already lost half of it. The act of explanation changes the dynamic. Readers try your strategy. The market responds. You are no longer competing against random noise. You are competing against every reader who implemented your idea. Rule two: alpha decays on a schedule that is roughly predictable but never exact. High-frequency strategies die in hours. Daily-rebalancing mean-reversion models can last months. Structural arbitrage between markets with different regulatory frameworks might last years. The longer your rebalancing period and the more dependent your edge is on market structure rather than pure statistics, the longer it will survive anonymity. Rule three: there is no way to publish and profit simultaneously. You can publish for reputation and lose the edge. You can keep it quiet and keep the edge. There is no third option that works consistently. Some people claim they publish abstractly while hiding the implementation details. This rarely works. Anyone with enough quantitative literacy can fill in the gaps, and the gap-fillers are usually the people who would have done it anyway.
When The Paradox Doesn't Apply
Not every published strategy dies immediately. There are classes of signals that persist because they are either too expensive to implement, too constrained by capital, or too dependent on conditions that cannot be generalized. For example, a strategy that requires a $50 million minimum position size to achieve meaningful returns after transaction costs will not attract copycats the same way a retail-friendly strategy will. A signal that only works during earnings announcement windows for a specific set of biotech stocks is hard to scale and harder to generalize. The most durable edges are structural. If your advantage comes from being physically located in a data center with lower latency to an exchange matching engine, publishing your code does nothing to help your competitors. They still need the colocation contract, which costs six figures and has a waiting list. If your edge is a proprietary data source that requires a multi-year contractual relationship with an exclusive data provider, the methodology section of your paper is useless to anyone who doesn't already have access to that data.
How To Test Whether Your Edge Will Survive Publication
Before you write anything public about a strategy, run this simple stress test. Describe your edge in a way that a competent quant could implement it without asking follow-up questions. If the description is under two pages and the implementation would take less than two weeks for a single developer, your edge is likely fragile. If the description requires thirty pages of caveats, depends on three proprietary data sources, and needs a team of four people to implement over three months, it might survive. Probably not completely, but it won't die overnight. Another practical test: ask yourself who else in the world has the infrastructure to trade this. If the answer is "anyone with a Bloomberg terminal and a Python environment," the edge is temporary. If the answer is "three people on the planet, one of whom is me," you have something more durable.
The Hard Truth About Publishing Research
The Sandy Sklarski Paradox is not a conspiracy. It is simply market efficiency working in real time. When an unknown analyst finds an edge and shares it, the market incorporates that knowledge faster than the analyst anticipated. This is not a flaw. It is the normal functioning of competitive markets. The analysts who last the longest are not the ones with the best signals. They are the ones who understand that their signals are temporary and who constantly move to the next one before the current one dies. Publishing is fine if you are publishing to build a career, not to protect an edge. But if you are holding capital that depends on that edge, keep it quiet. The market will find it eventually anyway. You might as well be the one still trading it when it does.
